Joint Research Centre

Transcription

1 Joint Research Centre the European Commission's in-house science service Serving society Stimulating innovation Supporting legislation Measuring the fiscal and equity impact of tax evasion in the EU: Evidence from Denmark and Estonia Sara Riscado 5 th International Conference on "The Shadow Economy, Tax Evasion and Informal Labour" Warsaw, July 2017

2 This is a joint work with Salvador Barrios (EC-JRC), Bent Greve and M. Azhar Hussain (Roskilde University), Alari Paulus (ISER-University of Essex) and Fidel Picos (EC- JRC). The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission. 26/7/2017 2

3 Outline 1. Motivation 2. Data 3. Estimation: methodology and results 4. Fiscal and distributional effects 5. Conclusion 26/7/2017 3

4 1.Motivation If VAT were paid by those who have to pay it, there would be no need to increase it so much Cristóbal Montoro, Spanish Minister of Finance, 9 th July 2012 Tax evasion o Shifts the tax burden from evaders to non-evaders, distorting consumption and labour supply decisions o Undermines the social contract between the state and the taxpayers o Weakens the redistributive nature of the tax and benefit system Effects especially relevant in times of severe economic crisis, when strong fiscal consolidation is required 26/7/2017 4

5 Our goal is threefold: 1.Motivation 1. To estimate individual measures of employment income underreporting 2. To measure the fiscal and distributional impact of eliminating tax non-compliance 3. To progressively extend EUROMOD coverage of tax evasion: o EUROMOD is a tax-benefit calculator for the 28 EU MS o Uses survey data from EU-SILC as input micro data o Need of accurate measures of true income o Up to now: Italy, Greece and Bulgaria can be corrected for accounting for non-compliance in a very aggregate way 26/7/2017 5

6 This paper Combination of micro-econometric and microsimulation approaches Use of administrative data together with survey data Focus on two countries: DK and EE Simulations using EUROMOD 26/7/2017 6

7 2. Data Quantification of tax evasion directly linked to the availability and quality of survey and administrative data: o EE: Exact respondent matching between SILC data and individual tax records, additionally pre-populated by third parties (employers) o DK: SILC data drawn from tax records need of complementary information (hidden economy surveys) and national aggregates 26/7/2017 7

8 2. Data: EE 2008 wave of national SILC for EE exactly matched with tax records o 14,942 individual observations in SILC, 99.5% linked with tax records complete employment information for 10,237 observations o No consent bias (no consent required to link datasets) o Information complemented with information from employers o Income from tax records is on average 87.5% of survey income (excluding zero earnings) 26/7/2017 8

9 2. Data: DK 2011 wave of national SILC for DK, based on administrative data + Cross-section studies on the hidden economy (Hvidtfeldt et al. 2010, Skov 2014, Skov et al. 2015) 'black activities' but also free exchanges of services: o Representative sample population aged between 18 and 74 o Covering the period , with a final total of respondents of in the final set o Includes also individual and household information on demographic, education, income, and labour market 26/7/2017 9

10 3. Estimation Methodology: EE o Estimation of true earnings/propensity to comply as a latent variable (Paulus, 2015) o Identification hypothesis: i. Civil servants do not lie, i.e. they cannot evade, so any difference between what they declare to the interviewer and the tax agency is due to measurement error ii. Measurement errors are distributed across all the sample, in the same way as they are distributed across the sub-sample of civil servants o Survey earnings-measurement error = true earnings =register earnings + non-reported earnings 26/7/

11 3. Estimation Methodology: EE True earnings Register earnings y i r = ln y T i = x i β T + ε T i, ε T 2 i N 0, σ T 0 if y T i = 0 (no earnings) 0 if y T i > 0 and r i 0 (full non compliance) r T i y i if y T i > 0 and 0 < r i < 1 (partial compliance) T y i if y T i > 0 and r i 1 (full compliance) Survey earnings ln y i s = θ s ln y i T 1 y i T > 0 + θ 0 s 1 y i T = 0 + x i β s + ε i s, ε i s N 0, σ s 2 Propensity to comply r i = θ r y i T + x i β r + ε i r, ε i r N 0, σ r 2 Estimation of the overall probability density function for a pair of observed individual earnings y i r, y i s conditional on true earnings through maximum likelihood ln L = ln f y i r, y i s 26/7/

12 3. Estimation Results: EE Estimated (true) status of employed individuals (%) Private employees All employees No earnings Fully non-compliant Partly compliant Fully compliant o Private employees: low full non-compliance around 4% and partial compliance around 30% o Including public employees: the fully and partially compliant groups drop to about 3% and 23% of the sample 26/7/

13 % 3. Estimation Results: EE Misreported earnings as a share of total gross true earnings (%) All Deciles of true earnings Tax non-compliance Measurement error o Substantial variation across the distribution of (true) earnings U- shaped profile o High variation in survey mismeasurement Tendency of people to present themselves more similar to the rest than they really are 26/7/

14 3. Estimation Methodology: DK o Tax evasion behaviour assumed to have three components: participation, number of hours per week, and the hourly wage rate o Regression equations dependent on individual characteristics, such as gender, age, family status, income levels, etc o For people participating in tax evasion activities, estimated non-reported income is obtained as the product between estimated hours spend and estimated wages earned on "hidden activities" 26/7/

15 3. Estimation Methodology: DK o o o o Decision to participate: logit regression and ranking of individuals (from highest to lowest) until reaching the national average in 2011 (Skov, 2014) Average number of weekly hours devoted to hidden activities, conditional on gender and age assigned to the individuals already found to participate in tax evading activities (Skov, 2014) Log of weekly wages: OLS regression fitted to the national tax evasion hourly wage in 2011 (Skov, 2014) For those participating: Non reported income = Hours per week charateristics evader wage rate charateristics evader 52 26/7/

16 3. Estimation Results: DK Estimated (true) status of individuals (%) Employees Whole population Fully non-compliant Partly compliant Fully compliant o More than two thirds of employees are fully compliant, around 24% engaged in hidden activities. o 6.2% of the population did not declare employment income but were involved in hidden activities (fully non-compliant) 26/7/

17 % 3. Estimation Results: DK Misreported earnings as a share of total gross true earnings (%) All Deciles of true positive earnings All employees Partially compliant employees o On average unreported income accounts for around 26% for partially compliant employees and 7% for employees. o Decreasing patterns on both graphs, except for the first decile of all employees where there are few evaders who hide big shares of their incomes. 26/7/

18 4. Fiscal and distributional effects Scenario Income Tax and benefits True baseline True Based on declared income No tax evasion True Based on true income 26/7/

19 4. Fiscal and distributional effects: EE Aggregate components of disposable income (million EUR) Tax evasion No tax evasion Difference Total Total Total Standard error 95% confidence interval Lower bound Upper bound % of baseline Original income 5,854 5, Taxes Social Insurance Contributions a Benefits 1,209 1, Disposable income 6,084 5, Inequality b a. Employees and self-employed b. Gini coefficient of equivalised disposable income. o Declared employment income increases taxes and social insurance contributions increase (13.9% and 8.9% respectively) and benefits go down (-0.8%) disposable income decreases (2.3%). o Inequality slightly decreases. 26/7/

20 Change as % of disposable income 4. Fiscal and distributional effects: EE Distributive impact of tax compliance on household disposable income (change as % of household disposable income) All Deciles of equivalised disposable income Reported original income Social Insurance Contributions (employee) Taxes Benefits Disposable income o 10% increase of reported income increase in taxes (2%) and SIC (less than 1%), no effect on benefits 2.3% reduction of disposable income. o Changes in disposable income mirror changes in taxes for all deciles. 26/7/

21 4. Fiscal and distributional effects: DK Aggregate components of disposable income (million EUR) Tax evasion No tax evasion Difference 95% confidence interval Standard % of Total Total Total Lower Upper error baseline bound bound Original income 1,029,445 1,029, Taxes 351, ,175 20, ,412 22, Social Insurance 91,728 97,553 5, ,413 6, Contributions a Benefits 312, ,677-4, ,836-2, Disposable income 898, ,393-31,162 1,188-33,491-28, Inequality b a. Employees and self-employed b. Gini coefficient of equivalised disposable income. o Effect have the same direction than in Estonia, but smaller relative changes in taxes (6%) and SIC (6.3%) have higher impact on disposable income (3.5%). o No significant change in inequality 26/7/

22 Change as % of disposable income 4. Fiscal and distributional effects: DK Distributive impact of tax compliance on household disposable income (change as % of household disposable income) All Deciles of equivalised disposable income Reported original income Social Insurance Contributions (employee) Taxes Benefits Disposable income o Increase in reported original income (7%) lower than in Estonia, but effect on disposable income higher (3.5%) due to the combined effect of the more significant effect on taxes, social contributions and benefits. o Taxes more relevant in higher deciles, benefits in lower. 26/7/

23 5. Conclusion Data availability conditions methodologies used and results obtained Linking survey and administrative data is fundamental for disentangling tax evasion and measurement error, as in EE Taking tax evasion into account is important to make accurate policy recommendations even in countries with high tax compliance, as in DK 26/7/