I-CUE Improving the Capacity and Usability of EUROMOD. Design Study implemented as a Specific Support Action

Transcription

1 I-CUE Improving the Capacity and Usability of EUROMOD Design Study implemented as a Specific Support Action Deliverable D3.8 I-CUE Feasibility Study: Lithuania October 2006 Contract number: Project Co-ordinator: Holly Sutherland Project website: Project funded by the European Community under the Structuring the European Research Area Specific Programme Research Infrastructures action

2 EUROMOD I-CUE FEASIBILITY STUDY I-CUE Feasibility Study LITHUANIA (2005 TAX-BENEFIT SYSTEM) Tatjana Stirling, Romas Lazutka October 2006

3 About I-CUE I-CUE (Improving the Capacity and Usability of EUROMOD) is a EUROMOD-related project that started in May 2005 and is supported by the FP6 Research Infrastructures Action as a Design Study. The aim of I-CUE is to re-design and up-grade EUROMOD in the light of: enlargement lessons learned from operating and using the first, prototype version. The main goals are to start the process of expanding EUROMOD to cover the 10 New Member States and to make EUROMOD easier to use, especially when it is dealing with 25 systems and datasets. This project involves the European Centre and the Institute for Social and Economic Research (ISER) at the University of Essex. The European Centre is responsible for establishing contacts and working relationships in the 10 New Member States in order to explore the feasibility of bringing them into EUROMOD. ISER is responsible for improving the model in a technical sense so that it is easier to use and to integrate the new countries. The main task of the Feasibility Studies is to lay the foundations for integration of the New Member States in EUROMOD, alongside the EU15, and therefore they all include: 1) key features of national tax-benefit systems; 2) identification of appropriate data requirements and data sources; 3) consideration of issues relevant for modelling each tax-benefit instrument (tax evasion, non take-up of benefits, etc.). For more information, see: and

4 I-CUE Feasibility Study LITHUANIA Tatjana Stirling, Romas Lazutka October 2006 Author information: Tatjana Stirling, Department for Work and Pensions, United Kingdom. Romas Lazutka, Department for Social Work, Vilnius University, and Institute for Social Research, Vilnius, Lithuania. 1

5 Table of Contents 1. INTRODUCTION AND OBJECTIVES THE MICRO-SIMULATION METHOD Model Output OVERVIEW OF LITHUANIA THE LITHUANIAN TAX AND BENEFIT SYSTEM Income Tax for an Individual Personal Income Tax: Expenses Deductible from Personal Income Taxation of Inherited Property Social Insurance Contributions Social insurance of self-employed people Payments to the Guarantee Fund Land Taxes Indirect Taxation Value Added Tax (VAT) Excise Taxes Social Security Social insurance system Medical insurance system Social support system INPUT DATASETS AND OTHER DATA REQUIREMENTS Data Requirements for the Simulation Input datasets Other Data Related Issues STEPS IN THE MICRO-SIMULATION PROCESS Particular modelling issues and ways to address them Sample Design Coverage Non-Response Changes in Macro-Economic, Demographic and Socio-Economic Areas Tax Evasion Non- Take Up of Benefits MAINTENANCE, SUPPORT AND TRAINING FINAL REMARKS SUMMARY OF TAX-BENEFIT SYSTEM RECOMMENDATIONS...40 REFERENCES...43 APPENDIX 1: Data from Lithuanian Government Departments

6 1. INTRODUCTION AND OBJECTIVES The new European Union countries will soon celebrate two years of membership. The past decade has brought a number of profound socio-economic changes. These changes have helped our countries get ready for accession to the EU. There is no doubt, however, that changes have multiplied since the day we joined. From a strategic point of view changes in financing conditions and a growing diversity of activities are the most important ones. Lithuania is one of the new member states. Analysis of financial conditions in Lithuania is a difficult task (as for other Central and Eastern European countries). This is mainly due to the lack of data. The only source of this type of information in Lithuania is the Department of Statistics. Public policy areas have to be expanded to include new issues and new problems both at the national level and in the EU domain soon after its expansion. The new member states will gradually adopt new practices of public administration and policy-making that are already established or have been recommended in the EU. Although the correlation of welfare and democracy is not direct, it is possible to note that the problems related with democracy satisfaction could be solved by public policy measures. Lithuanians who consider themselves as losers of transition reforms are unsatisfied with their socio-economic status and their anticipation of a prompt improvement in lifestyle. If such tendencies start to dominate in Lithuania, the question of political stability will arise. Therefore social security, welfare network and infrastructure development as well as the increase of the redistribution amounts, look like inevitable reforms in present-day Lithuania. All appropriate compensatory measures could be related to the development of the welfare state in Lithuania. The liberal model of the welfare state is the most suitable option for Lithuania at first sight. The Lisbon Strategy is implicitly based on this kind of welfare state model. Lithuania has tried to apply an active labour market policy but efforts were restrained by limited resources. The structural policy of the EU, mainly the measures financed by the European Social Fund, should increase the amount of funds allocated to the active labour market policy. When considering public policy in Lithuania, attention should be paid to the reduction of regional inequalities, establishment of the workplaces and social security infrastructure. Due to the Lithuanian budget restrictions it is necessary to develop a tax base and to improve tax administration. A possible tool is EUROMOD, a tax-benefit model for the EU currently covering the 15 old Member States. The purpose of the present feasibility study is to prepare the ground for developing a tool for modelling the effects of tax and benefit policy in Lithuania by extending and adapting a methodology that is already in use. The main idea is to identify difficulties, solutions 3

7 and prepare recommendations for building and developing a tax-benefit micro-simulation model for Lithuania. It is intended to take full account of relevant features of this country and to be compatible with the existing model for cross-country analysis. This is why we focus on the tax-benefit systems legally in force as of 30 June The report presents the Lithuanian tax-benefit system and discusses the feasibility of building a new country module for Lithuania within the European integrated micro-simulation model EUROMOD. Before we start, however, we need to address the following questions: Are there available resources to spend on preparing and building the model and will they be spent efficiently? Will the expected profit be higher than the expected cost? Essential information to build and maintain the model: 1. Detailed list of the specifics of Lithuanian polices in place. 2. Data requirement and availability. 3. IT software and knowledge. 4. Modelling techniques and difficulties implementing them. 5. Training and documentation for-policy makers. 2. THE MICRO-SIMULATION METHOD Micro-simulations are based on data derived from micro-databases containing information on individual, household and (regional) environmental attributes. The welfare impact of alternative economic policies, the incidence of taxes and benefits can be more accurately modelled on the basis of these micro-data than by means of aggregate functions. Reduced to its bare essentials, a data-based micro-simulation model suitable for this type of policy evaluations consists of two parts (Martini and Trivellato, 1997): a baseline database: a data set containing information on individual or family/household units, in particular socio-demographic characteristics and economic information related to a set of policies; a set of accounting rules: these are computer language instructions produced for each unit, provisions of existing or alternative tax and transfer systems, or other relevant institutional features. The construction of representative data sets containing all necessary variables and modelling at least part of a complex tax-benefit system absorbed all the resources in the early days of micro-simulation. The work of Pechman and Okner (1974) that analyses the redistributive effects of the US tax system represents the most significant example of this type of research. Generally, these models can be characterized as static as they simply reweight the dataset at each point in time to reflect the whole population. In addition, some 4

8 micro-simulation models incorporate a third component a set of behavioural relationships, which varies greatly in scope and importance across models. These can be of two types: behaviour that produces events which take place over time: demographic events, i.e. marriage, divorce, deaths, etc., and economic events: leaving the labour force; behaviour producing feedback reactions of individuals and/or families to changes of external circumstances, especially changes in public policies. Based on multivariate methods such as history event analysis or rule-based behavioural models, micro-simulation permits analysis of the interactions between variables and life course interactions between various parallel carriers and roles, such as education, work, partnership and parenthood within a changing socio-economic context. Demand for micro-simulation models in social security research does not only result from the special advantages of these models compared to others, but also from the fact that there is no alternative modelling strategy to address series of related critical policy and research issues. Caldwell and Morrison (2000) give the following examples: analysis of projected winners and losers on period-specific or lifetime basis; analysis focused on families and individuals simultaneously; exploration at the micro-level of the operation of social security programmes in the context of the broader tax/transfer system; quantifications of incentives to work, to save, or to retire at particular life course or period junctures; cross-subsidies across population segments or cohorts; feedback effects of government programmes on population demographics; and longer-term consequences of social trends in marriage, divorce and fertility. In response to the demands associated with prospective social security reform in the context of demographic change, decision-makers of various countries including the US, Canada and France have begun to use dynamic micro-simulation models to supply key policy inputs. Prominent examples of micro-simulation models used in the field of social security research and pension systems are CORSIM in the US, DYNACAN in Canada and DESTINIE in France. In CEE countries (in particular new entry countries) entirely new political regimes and economic reforms and legislations have come to power. At the same time, the demographic picture of the population has changed considerably (declining fertility and marriages, increasing life expectancy, etc.) together with migration to the old EU countries. To consider these issues in a more rational way it is essential to get tax-benefit models in place. At the present time decisions are made on an ad-hoc basis, which has a detrimental effect on future policy analysis. Tax-benefit micro-simulation models can answer what-if questions and thus contribute to a better understanding of alternative policy measures and their budgetary implications 5

9 as well as their consequences for economic and social policy aims such as distributional equity or incentive neutrality, poverty analysis and so on. This report will also contribute towards better understanding of the availability and quality of data sources in the country, not only for policy analysis. This might help other researchers in carrying out other projects, i.e. poverty, income distribution and social welfare analysis. But the biggest issue in building the model is data sources and their availability together with their reliability. Possible techniques to improve the sources include matching several data sources together such as administrative sources, census information and survey data. 2.1 Model Output. 1. The Tax-Benefit Model has two basic elements: a base dataset and the tax and income related benefit system translated into computer language (which could be C++, SAS or any other computer programming language). 2. The base dataset provides the micro data used within the model to represent the population and it is derived almost entirely from the latest or given year. Each record in the database represents a benefit unit and contains information about the unit in whole as well as information on each person in the unit. This includes a wide range of demographic and economic indicators as well as benefit unit and household structure. Each individual, benefit unit and household within the database can be uniquely identified. 3. The model itself will normally forecast a policy year later than that represented by the survey data. It is therefore necessary to uprate the data to make them representative for the appropriate policy year. In addition, like with any sample survey data, each benefit unit represents a large number of similar benefit units in the total population. This number is represented by the grossing factor supplied along with the used survey. In reality the grossing factors differ case by case as a result of adjustments to allow for relative non-response within different groups. 4. The computer code representing the tax and benefit system consists of a series of modules which simulate the policy rules of the income tax, social contributions and income related benefit systems. In almost all cases non-income related benefits are not modelled. Each benefit unit in the uprated dataset is run through this series of modules in order to model tax liabilities and income related benefit entitlements under the defined policy year system. The resultant dataset would be available as the base Model dataset, along with the various modules. 5. Future Model users would be able to make modifications to parameters or underlying codes of the model and create their own datasets under their alternative scenario and then compare it with the base. Subsequently it would be 6

10 possible to compare changes in liabilities, entitlements and final incomes on a case-by-case basis or as summarised results over specified groups of the population. This allows estimates of the distributional effects of policy changes to be simulated as well as of the net costs in terms of public sector revenue. Qualified model users familiar with the social and fiscal policy system in the country could recognise the parameters in the model and decide which ones they need to change. 3. OVERVIEW OF LITHUANIA In this feasibility study report we concentrate on Lithuania s socio-economic structure. These are some facts about Lithuania. 67% of Lithuania s 3.5 million citizens live in urban areas, giving the country a population density of 53.5 people per square km (138.6 people per sq. mi.). The population is 83.5% Lithuanian, 6.7% Polish, 6.3% Russian and 3.5% other (Belarusian, Ukrainian, Latvian, etc.). The official state language is Lithuanian, which is closely related to Sanskrit and belongs to the Baltic family of Indo-European languages. The 32-letter Lithuanian alphabet is Latin based. Main religion is Roman Catholic. Labour Force Total population of Lithuania 3.5 million. Lithuanian labour force 1.6 million; over two-thirds employed in the private sector. Lithuanian employee skills 42% with higher education, 24% with specialised education (i.e. technical certificates). Currency Lithuania s currency is the Litas (LTL), equal to 100 Lithuanian cents. Under a Currency Board system, the Litas is presently pegged to the Euro at a rate of :1. Under the Currency Board, the amount of currency in circulation is tied to reserves of the Bank of Lithuania. Economy Lithuania has one of the fastest growing economies in Central and Eastern Europe, with the private sector now producing more than 80% of the country s GDP. 7

11 Wages Lithuania s well-developed industrial base includes electronics, chemicals, machine tooling, metal processing, construction materials, food processing and light industry, including the manufacture of textiles, clothing, furniture and household appliances, and is complemented by strong transportation and service sectors. Lithuania s labour costs are among the lowest in the European Union. Since May 1 st, 2004, the minimum monthly salary is 500 LTL ( 145), the minimum hourly salary is 2.95 LTL ( 0.9). The average gross monthly wage in the 4 th quarter of 2004 was 1310 LTL ( 379). Table 1: Sector Average Monthly Salaries in Industrial Sectors, Q Salary Transport and Communications 1396LTL ( 404) Construction 1359LTL ( 394) Manufacturing 1224LTL ( 355) Source: Lithuanian Department of Statistics Table 2: Sector Highest Average Monthly Salaries by Sector, Q Salary Administration institutions 3248LTL ( 941) Financial intermediation 2640 LTL ( 765) Insurance and pension funding 2121 LTL ( 614) Source: Lithuanian Department of Statistics Table 3: Sector Lowest Average Monthly Salaries by Sector, Q Salary Agriculture, hunting and forestry 1022 LTL ( 296) Manufacture of wood and wood products 910 LTL ( 264) Hotels and restaurants 818 LTL ( 237) Source: Lithuanian Department of Statistics 8

12 Working Hours The regular working week is 40 hours. A shorter period may be negotiated. There is a national minimum annual holiday/vacation of 28 calendar days, not including public holidays. Maternity leave and childcare leave are possible until a child reaches the age of 3. Education Lithuania has the best-educated workforce in Central and Eastern Europe. According to the Lithuanian Department of Statistics, its proportion of graduates is the highest in CEE, with 4.3 university graduates per year per 1000 inhabitants. Tertiary Graduates 2003 Fields of Specialisation Figure1: In , the total number of tertiary graduates was Source: Lithuanian Department of Statistics. Most educational institutions are run by the state though several private gymnasiums, lyceums and other educational institutions (including private business schools) have recently been established. At present there are 21 universities and 27 colleges with a total enrolment of students. 9

13 4. THE LITHUANIAN TAX AND BENEFIT SYSTEM The financial and tax year coincides with the calendar year. However, a different tax year may be established taking into account the peculiarities of the taxpayer s activity. A taxpayer, upon the consent of the Tax Inspectorate, may have a varying 12-month tax year. This may be necessary due to the seasonal nature of activity or if the group to which the taxpayer belongs applies a tax year different from a calendar year. In this section, all personal taxes and benefits are listed from the simulation prospect. The tax benefit model supposes to include all personal tax reliabilities and benefits that an individual may be entitled to in order to get the best estimates. Therefore the model needs to be based on individual records. The data we are going to use is the Household Budget Survey. These are micro-level data with extensive records on income and expenditure. Therefore suggestions will be based on the data available. More information will be provided in section 5.1. The below presented tax-benefit rules reflect the system of direct and indirect taxes and benefits in the Lithuanian legislation as of June 30th, Income Tax for an Individual An individual in Lithuania is liable for tax on his/her income as an employee and on income as a self-employed person. For an individual who qualifies as a permanent resident of Lithuania, tax will be calculated on his/her income earned in Lithuania and on that earned overseas. A foreign resident pays tax only on his/her income in Lithuania. To be considered a Lithuanian resident an individual must meet the requirement of residence in Lithuania for at least 183 consecutive days in a 12 month period. Occasionally, an individual will be considered a Lithuanian resident even if he/she is a resident in Lithuania for less than 183 days if he/she owns a home in Lithuania that is his/her permanent residence. An employer is obligated to immediately deduct the amount of tax and national insurance due each month from a worker s salary. A salaried employee s income that is derived mainly from a Lithuanian company is taxable at 33%. Income tax in Lithuania consists of personal income tax and corporate income tax. In 2004 personal income tax amounted to 23.6% of the state budget revenue. Table 4 shows that in Lithuania, the budget revenue consists approximately to 86% of the taxes in the year Table 8 provides more detailed information on Government budget revenue and expenditure in year The majority of the direct tax revenue comes from personal income tax ( %), more than from any other direct taxation form. As to indirect taxes, the revenue is generated from the value added tax, about 34% from excises on alcohol, whereas tobacco and fuel make up for 15.5%. 10

14 Table 4: State and municipal budgets revenues by type of taxes (2003, %) Total revenue 100 Tax revenue 85.6 Personal income tax 23.5 Profit tax 6.9 Taxes on property 2.7 VAT 33.7 Excises 15.5 Taxes on international trade and transactions 2.3 Source: the Ministry of Finance Personal Income Tax: The Lithuanian taxation system is simple, there are different tax rates applying to different income sources. Personal income tax is one of the most important sources of income for the state in Lithuania. Each taxpayer is obligated to pay it, but they are entitled to a non-taxable minimum (NTM) on wages from their first (main) job. There are some types of personal exemptions: basic for all: 290 LTL ( 84). parents having 3 and more children under 18: 430 LTL ( 125); for the fourth and each subsequent child 46 LTL ( 13) in addition; parents having 1 or 2 children: 290 LTL ( 84) + 29 LTL (0.1 of the basic nontaxable minimum) ( 8.40) for each child; single parents having one child under 18: 335 LTL ( 97); for the second and each subsequent child NTM is increased by 53 LTL ( 15); for disabled of Group 1 up to 430 LTL ( 125); for disabled of Group 2 up to 380 LTL ( 110). The majority of social benefits is not subject to taxation. Exceptions are Sickness and Maternity/Paternity Benefits these are taxed according to the general wage taxation rules. It is possible to simulate tax exceptions, as there is information on a person s employment status, his/her wages in the data, and full information on dependant children. There is only a difficulty with a disabled person s NTM. There are no data about a person s disability status (group). Therefore we can use data variable on the paid personal income data table for a disabled person. 11

15 4.1.2 Expenses Deductible from Personal Income Cumulative life insurance premiums paid on the resident s own behalf, on behalf of his/her spouse and children; Pension contributions to pension funds on own behalf and on behalf of a spouse and minor children; Interest on housing construction/acquisition loans; Tuition (university education, acquisition of qualification, PhD, post-graduate studies), including tuition paid for children under 26. If a loan was taken for that purpose, the amount of loan returned over a tax period is deducted. This incentive may also be used by a foster parent, a brother, a sister or a spouse; One personal computer including software acquired within three years for an amount not exceeding 4000 LTL. If the purchase agreement contains a provision that title to the thing is transferred only subject to final settlement, only the amount actually paid excluding interest may be deductible; The total deducted amount may not exceed 25% of non-taxable income. Expenses are deducted only from a permanent resident s personal income while calculating income tax for the tax period for the purpose of submitting the annual income tax return. Can be simulated. There are data available on payments made by an individual. Only deduction with personal computer can be difficult to simulate as it needs historical data and we do not know whether it was deducted previously or not, therefore, this deduction needs to be ignored Taxation of Inherited Property Inheritance tax is applied to both Lithuanian and non-lithuanian residents (unless international treaties provide otherwise). The tax object of a Lithuanian permanent resident is inherited property - movable, immovable, securities and cash. The tax object of a temporary resident is inherited movable property requiring legal registration in Lithuania or immovable property located in Lithuania. The tariff of inheritance tax applied to inheritors is 5% when the taxable value is less than 0.5 million LTL (approx. EUR ) and 10% when the taxable value exceeds 0.5 million LTL. Close relatives: children, parents, spouses and certain other categories of people may be exempt from this tax. Inherited property valued below LTL (approx. EUR 2.896) is not subject to tax). Cannot be simulated. 12

16 4.2 Social Insurance Contributions Social insurance contributions are paid by employees and employers, the self-employed and farmers. In 2004 an employer was required to contribute 31 and an employee 3 per cent from gross earnings. People on social benefits do not pay social insurance contributions. Can be simulated or partially simulated. The main data source is the Household Budget Survey. Earnings are provided as net income, but there are separately recoded payments for social insurance contributions, therefore they are supposed to be easy to simulate Social insurance of self-employed people In Lithuania there are two categories of self-employed people. The first category covers those having a trade certificate (licence), the second category all others. The self-employed people are insured only for a basic social insurance pension if their monthly income is below the minimum wage. They pay social insurance contributions equal to 50% of basic pension. The second category of self-employed people provides insurance for a basic and earnings related pension if their monthly incomes are above the minimum wage. They pay social insurance contributions equal to 50% of the basic pension plus 15% of their income. Cannot be simulated. In the dataset there is no distinction between the categories of the self-employed. But there are data on the amounts paid for social insurance contributions. Therefore we will need to use data directly in our calculations Payments to the Guarantee Fund Enterprises with a residence in Lithuania, except branches, permanent establishments and representative offices in Lithuania, must make contributions to the Guarantee Fund: 0.2% of the employees gross salary is allocated to such contributions (which are the basis for calculating social insurance contributions). Can be simulated and is 0.2% of total social insurance contributions. Therefore it will be calculated parallel to the simulation of social insurance contributions. 13

17 4.3 Land Taxes Landowners pay Land tax. The annual tax rate amounts to 1.5% of the value of the land assessed in accordance with the Land Evaluation Methodology established by the Government Resolution No 205 of 24 February 1999 and applying the adjustment rations established by the Government. It varies from 1.5% to 6% of its value assessed in accordance with the Land Evaluation Methodology. Cannot be simulated, therefore the amount recoded in the data will be taken for the simulations. 4.4 Indirect Taxation Concerns regarding the relatively wide tax wedge have led to a planned gradual reduction of the personal income tax (PIT). Especially from 2008 onward, the revenue shortfalls from income tax cuts could reduce revenues by about 2 percentage points of GDP, jeopardizing the deficit target of 1 percent of GDP. The authorities plan projects that revenues especially from indirect taxes will partly offset losses from reduced income tax rates. Despite commendable progress made in tax administration, a continued increase in the revenue-to-gdp ratio cannot be presumed. The plan also projects wage restraint. On the revenue side, reducing exemptions associated with PIT and the value-added tax (VAT) should broaden the tax base and help improve administration Value Added Tax (VAT). Table 5: The rates of VAT in Lithuania are listed in the table below: Standard tariff 18 Reduced tariff 1st: 9 Building and renovation services financed by public resources (State and municipal budgets) Reduced tariff 2nd: 5 Passenger transport services Printing (books, newspapers, magazines) Medicine Accommodation (hotel) services Eco-friendly food Carcass meat, fowl, fish Services of agricultural cooperatives to own members A zero rate is applicable to export and transit of goods and related services. 14

18 4.4.2 Excise Taxes Excise taxes are paid for luxury goods and services like alcoholic beverages, tobacco products, cars, electronic equipment and gambling. The table below shows a list of excise rates for most common goods. Table 6: Excise tax rates Alcohol (per hectolitre) Beer Wine (alcohol concentration is below or equal 8.5%) Wine (alcohol concentration is above 8.5%) Alcohol products (alcohol concentration is below or equal to 15%) Alcohol products (alcohol concentration is above 15%) Ethyl alcohol Tobacco Cigarettes Cigars Smoking tobacco Benzene (per ton) Benzene without lead Benzene with lead Kerosene Gas * Euro/LTL=1/ LTL* per 1% of alcohol concentration 40 LTL 150 LTL 150 LTL 230 LTL 3200 LTL 47.5 LTL per kilo and 15% on price 38 LTL per kilo 111 LTL per kilo 1318 LTL 1934 LTL 1002 LTL 432 LTL Can be simulated if we plan to do so. The Household Expenditure Survey can be used to calculate household expenditure and indirect taxes. A reduction in the PIT rate in 2006 and proposals of changes to indirect taxation are the basis to consider modelling them. Euromod currently does not include indirect taxation, but it could be important for Lithuanian policy-makers to analyse interactions between the direct and indirect tax systems therefore we need to consider modelling it. 4.5 Social Security The Ministry of Social Protection and Labour of the Republic of Lithuania determines and implements policies regarding social support and social insurance, prepares drafts of legal acts and adopts legal acts within its competence. 15

19 The Board of the State Social Security Fund and its territorial departments collect contributions and distribute insurance payments. The Ministry of Health Care determines medical care policies, prepares drafts of legal acts and adopts legal acts within its authority. The Board of Compulsory Medical Insurance, an administrative unit under the Government of the Republic of Lithuania, and the National Patients Fund are the main institutions in charge of compulsory medical insurance. The National Patients Fund receives its funding from a portion of the social insurance contributions made to the State Social Security Fund and a portion of personal income taxes collected by the State Tax Inspectorate. The Securities Commission is the main supervisor of pension fund activities. In this respect, the Securities Commission is empowered to issue legal acts and regulations within its competence. Its main functions are to issue and revoke licences for pension fund activities, to grant permissions to reorganise pension funds, to provide official interpretations and recommendations on pension fund activities, to apply sanctions and to take other actions within its authority. The Social Security System in Lithuania is comprised of a social insurance system, medical insurance system and social support system. The most significant components of the State Social Security System are the Social Insurance System (including pension insurance) and Medical Insurance System. At present all employees are insured on a mandatory basis. Employers (i.e. Lithuanian economic entities or individuals that pay wages to employees) are required by law to pay or withhold contributions in full on behalf of their employees. Additionally, the law also provides for voluntary social insurance. Lithuania has a state social security system providing for compulsory social insurance for all permanent residents (employees and selected self-employed) in Lithuania and social assistance from state funds. Benefits provided by state social insurance include: o pensions o sickness allowances, maternity and child-birth benefits, child care benefits o unemployment benefits The total social security contribution is 34% from gross earnings (31% by the employer and 3% by the employee). Contributions to the Lithuanian Social Insurance Fund are fully tax deductible Social insurance system Social Security Contributions by Employers Employers pay a mandatory social security contribution for every employee to the State Social Security Fund. The employer s contribution equals 31% of the employee s gross wage: 1% labour accidents insurance, 3% medical insurance, 22.5% pension 16

20 insurance; 3% sickness and maternity or paternity insurance; and 1.5% unemployment insurance. Therefore, social security contributions depend on an employee s gross wage which may not be less than the minimum wage set by the Government (from 1 st January 2003 it is set at 430 LTL (EUR ) per month). The employer s social security contributions are not deducted from gross wage, but must be calculated separately on top of the gross salary and paid not later than the last working day before the 15th day of the following month. The State Social Security Fund transfers medical insurance contributions to the National Patients Fund. Pensions The current Pension Law came into force on January 1, It provides three types of pensions: old age, disability, and survivor/widow(er) pensions. Pensions are calculated according to a set of formulas taking into account salary and years of service. Lithuania is reforming its retirement pension system. Old-age pensions are being shifted to a scheme based on years worked and the amount paid into the state pension fund. Currently, women who have reached the age of 59 and men who have reached the age of 62 and 6 months are eligible for old-age pensions (in 2006, the retirement age will increase to 60 for women and 62.5 years for men). The Insurance System is based on pay-as-you-go financing that provides a pension with a flat-rate (basic pension) and an earnings related element (supplementary pension). From January 1 st, 2004, employees can allocate a portion of their social security contribution equal to 2.5% of total income to an individual account. The portion of the social security contribution, which can be allocated, will increase by 1% a year until it reaches 5.5% in Early pension: Up to 5 years before the normal retirement age with 30 years of insurance and after 1 year of unemployment (introduced from July 1 st, 2004). Old-Age Pension Minimum period of membership: 15 years of pension insurance. Conditions for drawing a full pension: 30 years. Legal retirement age: Men: 62.5 years; Women 59.5 years in 2005, 60 years in Determining factors of old-age benefits: insured income (wage and short-term social insurance benefits) and number of insured years. The monthly Old-Age Pension is calculated according to the formula: P = B + S B: basic pension determined by the Government that cannot be less than 110% of the Minimum Standard of Living (MSL equals to 125 LTL or 36 which is the basic amount for the calculation of most of the categorical social benefits 17

21 respectively to 185 LTL or 53 which is the benchmark for means-tested social assistance). Since September 2005, the basic pension is 200 LTL. S: supplementary pension, 0.5% of the wage earned in each year. Cannot be simulated due to a pension s dependence on historical data, which are not available. The amounts will be directly taken from the data. Disability Pension Disability pension: The pension varies according to the assessed degree of disability. Group 1: formula for calculation of the Old-age Pension plus supplement equal to 50 per cent of basic pension, Group 2: formula for calculation of the Old-age Pension, Group 3: Only half of the Disability Pension for Group 2 is paid to Group 3 invalids Disability pension: Disability involving either a permanent or prolonged incapacity to work. Depending on age, the insured person must have been working for a minimal term in order to build a sufficient social insurance work record. Cannot be simulated due to a pension s dependence on historical data that is not available. The disability status is also not known. Survivor Pension Survivor pension: Paid to a spouse who has reached old age or is disabled. The spouse receives 20% of the insured person s pension. Orphan s pension: Orphans up to age 18 (age 24, if a student) each receive 25% of the insured person s pension. The total survivor pension must not exceed 80% of the insured person s pension. Survivor pension: The insured person must have been a pensioner or have been entitled to a disability pension at the time of death. Widows or widowers who were widowed before January 1 st, 1995 (when new legislation came into force) receive 25% of basic pension. Cannot be simulated due to dependence on historical data and information on other family members (that used to be the main provider but now he/she is dead) that is not available. 18

22 2nd pillar Old-age Pension Scheme A person insured for full pension insurance (basic and supplementary parts of pension) may voluntarily choose either to stay only in the social insurance system or switch to the 2nd pillar and direct a part of social insurance contributions, dedicated for the supplementary part of old age pension, to a personal account in a chosen privately managed pension fund. Part of the contributions are directed to the pension funds (2.5% in 2004, 3.5% in 2005; 4.5% in 2006, 5.5% in 2007 and further). Cannot be simulated due to a pension s dependence on historical data that is not available and there is a need to make an assumption who is contributing to this system as it is voluntarily. And as its accumulations started in 2004, there are no beneficiaries, yet this would be relevant to model in a dynamic micro-simulation model, but now we are talking about creating a static micro-simulation model. Sickness or Temporary Disability Benefits A person must have social insurance coverage for at least 3 months during the past 12 months or 6 months during the past 24 months in order to get this benefit. An employer pays at least 80% (but not more than 100%) of the employee s Compensatory Wage for the first two days. After the first two days the Social Insurance Fund pays 85% of the average monthly Compensatory Wage. The compensatory wage cannot exceed 3.5 times the national average insured income. The benefit must not be lower than ¼ of the average insured income of the current year. Maternity or Paternity Benefits An initial period of benefit payment is made to a mother (Maternity Benefit) and then an extended benefit is payable either to a mother or a father (Maternity/Paternity Benefit). Maternity Benefit: Full Compensatory Wage is paid monthly for 70 calendar days preceding delivery and 56 days after delivery. Maternity/Paternity Benefit: This benefit, 70% of the Compensatory Wage, is payable after maternity leave has expired until the first birthday of the child to a parent who nurses a child. It must not be lower then 1/3 of the average insured income of the current year. s These benefits cannot be simulated since they are calculated on the basis of insured people s last earnings and their history of previous contributions and duration of different benefits, which can vary, but it is still possible to make qualitative assumptions or we can directly take received amounts from the data. 19

23 Unemployment Benefit Compulsory insurance covers individuals receiving remuneration for work. The required service record is at least 18 months during the last three years. Duration of payment. Unemployment benefits may be paid for 6 months if the people s service record is less than 25 years, 7 months if their service record is years, 8 months if their service record is years, and 9 months in the case of a service record of 35 years and over. Amount of the benefit. The unemployment benefit comprises fixed and variable components. The fixed component represents state-supported income (SSI, 135 LTL), and the variable component is equal to 40 per cent of the former wage. The ceiling is 70 per cent of the people s insured income. A person without the required service record can receive only the fixed component of the unemployment benefit. A training stipend for unemployed people is 150 per cent of MSL (187.5 LTL). Since these benefits are calculated on the basis of insured people s last earnings and their contribution history and duration of the different benefits can vary, it is not possible to simulate. It will be taken directly from the data. Temporary Disability Benefits Periodical Cash benefit. They are paid until recovery or determination of disability. The amount of benefit is 100% of the average monthly Compensatory Wage. The monthly compensatory wage comprises an average wage for the last calendar quarter before sickness, from which contributions to sickness and maternity insurance have been collected. Lump sum compensation for lost capacity to work is paid in the amount of: less than 20% loss of capacity: 10% of 24 times the monthly compensatory wage. The amount is 3 times higher in case of permanent incapacity; 20% to 30% loss of capacity: 20% of 24 times the monthly compensatory wage. The amount is 3 times higher in case of permanent incapacity. Cannot be simulated, as it would require previous wage records that are not available in the dataset. 20

24 Permanent incapacity The minimum level of incapacity giving entitlement to compensation is 30%. Periodical compensation of lost capacity to work is paid monthly. It is calculated according to the formula: 0.5*d*k*D where: d: lost capacity coefficient; k: compensation coefficient (ratio between an insured person s average monthly income from 12 consecutive months prior to injury and the current year s insured income which is valid at the time of injury); 0.25<k<3; D: current year s insured monthly income valid in the month of payment. Cannot be simulated, since there is no information on historical labour participation records. Therefore payable amount will be taken into simulations. Funeral Benefits. A periodical benefit for widows, widowers and orphans. Each recipient is given a benefit equal to a periodical compensation of lost capacity that the deceased would have received divided by the number of recipients plus one, e.g. if there were 4 entitled people they would each receive 1/5 of the deceased person s Disability Pension. It is paid on top of other benefits. Capital sum on death. A lump sum equal to100 times the monthly average wage in the country. It is divided equally between the family members of the deceased. Cannot be simulated. There are no data available on household members death and their employment income. But there are data on amounts of benefit received Medical insurance system Medical services are provided directly by government health facilities. Benefits include the cost of medication for inpatient treatment. Cost sharing: Part of the costs of medication for outpatient treatment for insured people is covered by the state social insurance budget. Cannot be simulated because it would be impossible to estimate. 21

25 4.5.3 Social support system In this context, a non-contributory family benefit is defined as a universal system financed by general revenues, which provides flat-rate (categorical) benefits. Birth Grant Family benefits A Birth Grant is paid to one of the parents or guardians raising a child. The grant for each child equals six times the MSL. Child Benefit Who can get the Child Benefit: Families with child/children aged 0-3 years old that have no right to Maternity/Paternity Benefit. Families with child/children aged 1-3 years old that are entitled to Maternity/Paternity Benefit. All families with child/children aged 3-6 years old. There are no conditions relating to periods of residence or employment. Means testing is not required. Monthly amounts 0-3 years: the benefit equals 75% of the MSL, which is fixed by the Government and equal to 125 LTL ( 36) per month. There is no variation by income. 3-6 years: 50 LTL Benefit for Families with Three or More Children Families with three or more children under 16 years old, or older than 16, if the child attends school or is a full-time student, receive an additional benefit. The benefit is equal to the MSL per month and is increased by 0.3 MSL for the fourth and any subsequent children. Birth Grant and Child Benefit can be simulated. There are enough data to do so. Maternity (Pregnancy) Benefit for Studying Women Maternity, or Pregnancy, Benefit is paid to a pregnant woman that is not a full-time student. The benefit amounts to 75% of the MSL for 70 days preceding delivery. Maternity Benefit cannot be simulated, it will be taken directly from the data. 22

26 Benefit for Children of Military Conscripts During times of compulsory military service a conscript receives a benefit of 150% of the MSL per month per child. Benefit for children of military conscripts cannot be simulated but there are data available from HBS. Orphan s stipend The Orphan s Stipend is paid to orphans enrolled in higher, tertiary or vocational school regardless of other stipends. The monthly amount is four times the MSL minus any Orphan s Pension. The Orphan s Stipend cannot be simulated; amounts of a benefit will be taken from the data. Foster Child Benefit. The Foster Child Benefit is paid to a person or a non-state care institution raising children, if they do not receive support from the State. The amount of benefit is equal to four times the MSL for each orphan or a fostered child up to the age of 18 years. If the child attends school after reaching 18 the benefit is continued. If an Orphan s Pension and/or alimony is paid for a fostered child, the size of the benefit is equal to the difference between the MSL and these benefits. Foster Child Benefit can be simulated; there is a relationship variable that allows us to identify these children. General Means-tested Social Benefit Minimum Income Guarantee A guaranteed minimum amount is paid, if there is lack of income (incomes below the State Supported Income of 135 LTL [ 39] per person per month). A family income (wages, pensions, family benefits, unemployment benefit, alimony, income from farming, etc) is taken into account when the Social Benefit is calculated. The guaranteed minimum amount depends on income per family member. The monthly benefit level is 90% of the difference between the actual family income and the State Supported Income. The benefit is granted for three months. It may be renewed an unlimited number of times if circumstances have not changed. 23

27 Indexation. Benefits are adjusted at irregular intervals according to a governmental decision based upon the price index. The general means-tested Social Benefit can be simulated as there is a vast amount of income data provided in the HBS. Housing A reimbursement for the cost of flat heating and water supply is provided for low-income families and is based upon a means test. A family should not have to pay more than 25% of a family income above 90% of the state supported income, i.e LTL ( 35) per family member for heating of a standard size accommodation; 5% of a family income for a basic standard of hot water; 2% of a family income for a basic standard of cold water. The housing benefit cannot be simulated because an estimation of standard size accommodation is not known. There are different standard sizes depending on local authorities. However it is possible to get amounts of housing benefit from the data. Social pension (Disability and Old-Age) A non-contributory social pension is paid without means testing to disabled (I and II group) people who are not entitled to receive social insurance old-age pensions and to people of retirement age who due to objective reasons are not entitled to receive social insurance old-age pensions. The amount depends on the social insurance basic pension fixed by the Government and a coefficient that can vary between 0.75 and 2 for different groups of disabled recipients. The coefficient for the elderly is 1. Social Pension cannot be simulated. As we have no way to identify these people, the amount will be taken from the data. Care Benefit for Totally Disabled People Totally disabled people or those taking care of them receive a Care Benefit. The benefit is equal to 100% of a social insurance basic pension. It is paid on top of a Disability Pension. Care benefit for totally disabled people cannot be modeled as there is no information on disability status. 24

28 Transport Compensation Transport compensation is paid at 25% of the monthly MSL to disabled people with mobility dysfunctions. Every six years disabled people who drive themselves in special cars are paid three lots of compensation at twice the MSL each time. Disabled people are also entitled to an 80% or 50% discount on railway, ship and bus tickets. This is dependent on their disability group and is covered by the State and in certain cases by local budgets. Transport compensation cannot be simulated. But the amounts are recorded in the data. Death Grant. The Death Grant is a universal, non-contributory grant of 750 LTL ( 217) paid in the case of death of a permanent resident. The grant is equal to six times the Minimum Standard of Living fixed by the Government Death Grant can not be simulated. No data available on dead family members. We can only use the available benefit amount. 5. INPUT DATASETS AND OTHER DATA REQUIREMENTS 5.1 Data Requirements for the Simulation In order to simulate or to start building a model it is necessary to create a wish-list of variables. These variables are required by the model and it is important to underline from which data sources they are taken. Mainly these simulations would be based on the Household Budget Survey. The Office of National Statistics of Lithuania is using the HBS for social-economic statistics. A standard wish list of variables is necessary for tax-benefit models. (For more information refer to Development of a Tax-Benefit Micro-simulation Model for Moldova by H. Immervoll and G. Tarcali, 2002): Demographic information requirements o General demographic information (age, gender, educational level achieved, current educational status) o Information on a household structure in order to identify relations within a family. This information is necessary for building different fiscal units for 25

29 a model (spouse, children, and parents). It would be available from a data source. Labour market information requirements Income Expenditures Expenditures o Employment status, working hours, industry, sector, occupation, civil servant status, degree of disability. In addition to available variables in the HBS we need to include information on war participation. o income from self-employment, agriculture, capital, rent, other private sources, public benefits and any other income that is relevant for taxes and benefits in Lithuania. The level of desegregation would be considered during the simulation work. o Since the tax system of Lithuania is based on individuals, all income components should be available for simulation at an individual level. o Detailed expenditure variables for goods that are subject to indirect taxation. Both data on quantity and price of different goods and services are necessary. o Detailed expenditure variables for goods that are subject to indirect taxation. Both data on quantity and price of different goods and services are necessary. 5.2 Input datasets There are several micro level datasets available in Lithuania that can be used in microsimulations: the Labour Force Survey, the Household Budget Survey and the Family Fertility Survey. Therefore we need to identify the best-suited source for our excises. Other household surveys are also reviewed. Sampling designs, comments on non-response and some main estimates, variables needed for the simulations are also discussed and on the bases of these comparisons we recommend the best source. The main dataset in Lithuania that can be used for micro-simulation purposes is the Household Budget Survey. It is widely used by the Lithuanian statistical department in order to calculate socio-economic characteristics. These are used to produce national statistics (see Table 10 in the appendix). 26

30 The Household Budget Survey (HBS), conducted in line with the current methodology, has been carried out since The methodology for the survey was created in 1995 with the aid of World Bank experts and upholding the Eurostat recommendations. The methodology of the survey was revised and improved in The Household Budget Survey is one of the most complicated statistical surveys conducted by Statistics Lithuania. The objective of the survey is to obtain reliable information about the standard of living of households: income, expenditure, household structure, housing conditions as well as acquisition of durable goods, etc. The target population of the HBS is based on private households in Lithuania. People living in institutional households such as nursing homes for elderly people, imprisonment institutions, compulsory military service installations etc. have been excluded from the current survey. Households are selected using the random sampling method and using the Population Register. Such sampling methodology is expected to ensure equal possibilities for the representatives of all social strata to be selected for the survey. The selected households are surveyed for a period of one month. After one month other households replace them. Each household participating in the survey is guaranteed anonymity of the information submitted to statistical agencies, based on article 15, Law on Statistics of the Republic of Lithuania. The HBS uses two data collection methods combined into one: an interview conducted by an interviewer and self-registration of particular household indicators (monetary expenditures, foodstuffs, tangible goods and services received free of charge as indicators of farming activity) in special survey paper sheets. There are different categories of monetary variables listed below. Disposable income (net income) is the total income from work in cash and in kind (income and social insurance tax excluded), income from farming, business, independent professional work, social benefits, property income, rent, regular support from other persons. Income from employment is wages and salaries in cash and kind, bonuses, fringe benefits, daily allowances for business trips, different compensations paid by an employer, etc. Income from self-employment (non-agricultural economic activity) is income from business, handicrafts, free professional activity and other activities. Calculations of income from self-employment include only net income, i.e. current expenses for raw and other materials, energy, equipment, business related taxes etc. are deducted from the received income. Households engaged in business activity are requested to indicate only that part of income allocated solely to household needs (including savings). Income from agriculture income in cash and kind received from agricultural production. Income in cash is calculated deducting expenses for agricultural production from the income received from sales of agricultural products. Since the expenses for agricultural production might exceed income from sales of agricultural products (during winter and spring in particular) income in cash from agricultural production might be less than 0, i.e. with a minus. Income in kind from agricultural production includes total agricultural production that is consumed by a household itself (fresh or processed in the household). Income from property interest received for bank deposits 27

31 and money lending to private persons, government securities (treasury bills) as well as dividends received by shareholders. Rent income received by households from renting immovable property (a house, garage etc.), land or other property. Social transfers social benefits made for the following: old age, in case of sickness, disabled people, survivors, family and children, unemployed, in cases of social exclusion. Other income scholarships, alimonies, money received from parents, children, relatives or friends, produced food products, received free of charge from other households. Consumption expenditure is expenditure in cash and in kind, aimed at satisfying household s consumer needs, i.e. expenditure on food, clothes, footwear, maintenance of dwelling, transport, health care, cultural, recreational and other needs. All disagregations are available from this data set that is very important for the tax benefit micro-simulation model. Datasets such as the Labour Force Survey could be useful for the simulation. The Labour Force Survey is the most reliable way to estimate the situation and changes in a labour market. The survey has been conducted in accordance with the European Council and European Commissions Regulations. People aged 15 and older, selected from the Population Register by the random sample method are interviewed. This approach ensures equal conditions for everybody of the surveyed age to be included in the survey. Eurostat requirements were taken into account for the preparation of the questionnaire. Employment and unemployment rates are internationally comparable because of these data. There are 4000 households selected quarterly for the continuous Labour Force Survey. Each quarter 25 per cent of the sampled households are replaced. The sample includes the population of all cities, towns and some villages. The survey is carried out by interviewers. Approximately 0.4 per cent of the population aged 15 and older is interviewed. The labour force survey does not record as many monetary variables as the HBS that is therefore the best source so far. Also, the HBS has bigger household numbers. Therefore it is our preferable data set for the tax-benefit model. The third set of data is the Lithuanian Family Fertility Survey (FFS). This was conducted on two sub-samples: female and male, aged 18-49, permanently living in Lithuania. The sampling procedure used was a multi-stage random, with random route methodology at the final stage, and Kish (1965) tables for the selection of the respondents. The interviewers visited 7,463 households in 131 settlements in Lithuania; 187 interviewers worked at 500 sampling points in total during the study. The total number of completed interviews was 5,000, with 3,000 females and 2,000 males. The Baltic Surveys Ltd. conducted the fieldwork. This survey would be very useful for the calculations but its sample size is quite small, and there is not enough information on income and no information on expenditure. Also it was conducted only once in 1994/1995. But there are historical records, but not on Labour participation. Therefore, the HBS still is the leader for the model-building exercises. 28

32 Official tax and social insurance records can be helpful in a situation where it is not possible to simulate some tax-benefit system elements. Also these data sources can be used for validating a model as well as distributional, or proportional, aspects of respondents in micro-level data. But administrative data sources are not available. Therefore, for the validation we can compare the results from the simulation with the recorded and look how well our model is doing, or alternatively compare with Government expenditure on the calculated benefit. 5.3 Other Data Related Issues Definition of Base Year of the Simulation Normally simulations are done on the latest data available or the given year of the data. It was agreed to do simulations for 2004 for the cross-country comparison purposes. Net-to-Gross Conversion Gross incomes are required as a basis for the calculations when simulating taxes and benefits. However, survey data frequently record some/all income components net of taxes and/or contributions. This implies that gross incomes should be imputed. In the case of Lithuania the HBS contains both net incomes and taxes paid so it is possible to recover gross incomes. At first, the quality of gross and net income data should be assessed. If the quality is adequate then net incomes plus taxes and contributions can simply be used for the simulation. If the quality of tax and contributions data is insufficient (and in cases where the relevant variables are missing), alternative approaches are necessary to impute gross incomes. In practice there are two possibilities for getting gross income from the available net. The first is the use of existing net-gross tables that exactly describe gross values for different net amounts. The second is to use a micro-simulation model to reverse-simulate gross incomes from net. The idea here is to exploit an existing model of detailed taxes and contributions and apply it in an iterative way to different estimates of gross income. For each estimate, simulated taxes and contributions are deducted to arrive at a simulated net income. If this simulated value is different from the net income recorded in the data then the estimated gross income is adjusted and the entire process is repeated again until an acceptable match of simulated and recorded net income is found. This procedure has already been successfully used in other countries (refer to Immervoll and O Donoghue, 2001a for details). The Lithuanian Office of National Statistics is collecting net incomes, as well as social insurance contributions and income tax, so the sum of these will provide the gross income required for the simulations. Also, the Lithuanian tax system has flat tax rate, and tax amounts are corresponding to the recorded value. Furthermore, it is supposed to add up as there are no tax credits in Lithuania that could compromise this approach. 29

33 Data Cleaning Since tax-benefit models have a deterministic nature, they require input data that do not have any missing values for variables required for the computation of simulated taxbenefit instruments. As a result, it is necessary to thoroughly check the degree of units missing, non-response and item-non-response and to take appropriate measures to impute missing information. 6. STEPS IN THE MICRO-SIMULATION PROCESS Micro-simulation techniques used by tax-benefit models are universal. The structure of modules is shown in Figure 3. The individual components shown are related as follows. The data accesses elements separately, reads information on each micro-entity from the micro-database and updates monetary information. All observations (e.g. individuals) that are part of this micro-entity (e.g. household) are then grouped together to form the assessment units to which the tax-benefit rules relate. The core of the model consists of the tax-benefit algorithms. For each instrument, these are performed separately for each assessment unit. The algorithms are organised according to the generalised structure of tax-benefit systems discussed above and are evoked in the appropriate sequence. All outputs of these calculations (monetary amounts of the simulated instruments, measures of disposable income, etc.) are then written to a micro-output file containing all variables of interest for each unit of analysis (e.g., individual or household). 1 This main loop, indicated in figure 3 by bold boxes and connectors, is repeated over all (N) micro-entities stored in the micro-data. This approach is common across all counties which are using micro-simulation modelling techniques for their tax benefit model; however, there will always be some elements which are unique to the country of interest. These are tax benefit rules specific to the country. The implementation of the various model components requires large amounts of expertise, effort, time and resources. It is therefore useful to base any new model-building exercise on work that has already been done in the context of similar projects. We propose to use, wherever possible, an existing model-building framework such as XLSIM (see O Donoghue). It is an Excel-based interface which is easy to use. A model created in the XLSIM model framework should not be difficult to transfer to C++ Euromod programming language. The Lithuanian Tax-Benefit system is relatively simple, therefore the model would not require long in coding Lithuanian rules. 1 The unit of analysis (for instance the household for measuring poverty) will often be different from the various units of assessment (e.g., the married couple in the case of joint income taxation) required for computing tax and benefit amounts. 30

34 Figure 3: Basic Components of a tax-benefit micro-simulation model. Source: Immervoll and O Donoghue, 2001b. 31

35 6.1 Particular modelling issues and ways to address them Due to non-existence of tax-benefit policy modelling procedures, all decisions in Lithuania are made on an ad-hoc basis. So building a model for this purpose will give a greater benefit than costs involved. But it is necessary to consider specific country policies and data issues. It is important to get acquainted with them and make the best possible solution scenarios. Any model can only be as accurate as the data it is based on. So issues of data quality and representativeness of the population can and are causing big concern in the modelling world. The non-response and under-reporting of the information can be a problem. This chapter attempts to give recommendations how to deal with these issues in the modelling world Sample Design The target population consists of all private households in Lithuania households were selected in HBS for the 2004 survey of which 7961 participated in the survey. The Population Register was used as a sampling frame. A stratified sample design with a simple random sample and a two-stage cluster sample was used. Lithuania was divided into 31 not overlapping groups strata. The biggest cities of Lithuanian counties, medium, small towns and rural areas of counties were divided in separate strata. A sample of households was selected from each stratum. A different sample design was used within each stratum. A simple random sample of people 16 and older is drawn from the Population Register in major county towns. Only the households pertaining to the address of the selected people are under the scope of the survey. According to this method, 4794 households were selected in the biggest cities of the counties. A two-stage cluster sample design was used in medium and small towns of counties. The Pareto sample with probability proportional to the cluster size was used in the first stage. Each town is a cluster. A simple random sampling of people aged 16 and older withdrawn from the Population Register was used in already selected clusters at the second stage. Households living at the selected addresses were surveyed. The approach applied to medium and small towns was also used in rural areas of the counties. Though the design of a sample is supposed to represent the whole country and its counties, due to the fact that samples are small for the counties, just key indicators are calculated since these have the smallest areas associated with them. 32

36 6.1.2 Coverage Surveys usually do not cover the whole population. People living in institutional households (nursing homes for elderly people, imprisonment institutions, compulsory military service installations, etc.) have been excluded from the current survey. Therefore results of models based exclusively on these data exclude these fractions of the population, which is necessary to take into account when interpreting a model output such as poverty rates, inequality measures or total revenues/expenditures. If the exclusion of potential observation units is complete and systematic (i.e., by region, occupation or employment status) then traditional survey design techniques (weighting) cannot be used to ensure that the resulting sample is representative. In order to improve the model it is necessary to get access to additional sources of information: census, small sample area statistics and/or administration data. The later could be provided by Lithuanian Government departments and in this case the model would benefit the most. Also the aim of this work is to show the benefits of this modelling to the Lithuanian Government Department that could actually benefit from the administrative data to which they have access Non-Response In every statistical or sociological survey it is important to have as many sampled respondents as possible to answer the questions. It is directly connected to the accuracy of the results gained. However, experience in statistical surveys shows that it is actually impossible to interview all respondents. There are many reasons why households may not respond to surveys. It may happen that interviewers fail to find the household required or the member of the household found may refuse or be unable to participate in the survey. Some of the households are not interviewed due to inaccuracy of data in the Population Register. It was observed that the greatest number of respondents not interviewed occurred in the largest cities (41 %) and the least number in rural areas (13.4%). One of the main reasons was their refusal to participate in the survey. In the largest cities even 29% of all selected households refused to participate, in other towns 14% and in rural areas 7%. The causes of their refusals are rather diverse. Some of them are afraid to give personal information, others are not interested in the survey or do not trust the authorities, or refuse due to a lack of time. 33

37 Table 7: Non-response rates by reasons, per cent Total nonresponse refusals absence other reasons Total 26,7 17,2 4,8 4,8 I Q 25,0 15,9 4,5 3,6 II Q 26,4 17,0 4,9 4,5 III Q 29,2 17,5 6,6 5,1 /V Q 27,3 18,1 3,2 6,1 5 largest 41,0 28,6 7,4 4,9 cities 1 Q 37,1 27,1 6,2 3,9 11 Q 39,4 26,6 7,9 4,8 III Q 44,6 29,4 10,9 4,3 IV Q 42,3 31,1 4,7 6,5 Other towns 23,8 14,5 5,6 3,8 1 Q 22,7 14,1 5,9 2,7 II Q 25,9 16,4 5,5 4,1 111 Q 24,7 13,3 7,2 4,2 /V Q 21,9 14,1 3,8 4,1 Rural areas 13,4 6,8 1,0 5,7 1 Q 11,1 5,7 1,2 4,2 II Q 12,8 7,2 1,1 4,5 III Q 15,2 7,5 0,9 6,7 IV Q 14,7 6,6 0,8 7,3 Not all the households selected for the sample participate in the survey due to refusal or other reasons. Therefore parameters were estimated using the sample design weights and data from households that participated in the survey. The sample design weights were calibrated to reduce bias whilst remaining as close as possible to the original sample weights. This also allowed certain demographic estimates such as number of people in each stratum, number of people by gender and some age groups to be as close to reality as possible. To build our model and conduct analysis it is important to use weighted survey data. It would take account of the non-response issue. Weights would make our data representative on population level. In order to make it representative for future years we can use other sources of information available on aggregate level (administrative data, population estimates, etc.) and gross up the data to the control totals Changes in Macro-Economic, Demographic and Socio-Economic Areas Usually data on micro-level (surveys) is available only one to two years after it has been collected. We are experiencing the same situation with Lithuanian data nowadays. 34

38 However, we would be interested in producing model results that would output datasets a few years in advance in order to estimate policy effects or alternative policy scenarios. Therefore it is necessary to find a way to project the data forwards. It means all economic characteristics have to be adjusted to the year of the simulation needed. This process of projecting micro data forward is called ageing or uprating. Depending on circumstances, data are projected forward to reflect people s ageing to the year of simulation, therefore we are aging the population in our module. Another possible way is to leave age untouched and only uprate income-related characteristics to uprate the data. Adjustments may also be required for forecasting purposes. In this case the underlying data would be adjusted to match expected changes in the population and incomes. Data thus adjusted can then be used as input into the tax-benefit model to explore projected aggregate revenue/costs and/or distributional features. These issues are well-known and there are a number of techniques to address them. However, as the term transition implies, the speed of macro-economic, demographic and socio-economic changes in transition economies may render micro-data sources obsolete much more quickly than in developed countries. The Lithuanian economy has recently stabilised and it cannot be considered a transition country. Relevant dimensions include changes in price levels, level or structure of economic activity and changes in unemployment. Since Lithuania has become part of the EU a new labour market has appeared. The Lithuanian labour force started to migrate to the old EU countries in large numbers which influenced the Lithuanian economy. Therefore this issue needs to be addressed in the Lithuanian tax-benefit building project plan. The first type of adjustment technique relates to the value of monetary variables by simply applying uprating factors that approximate the change in a value since the data were collected. For instance, indices are usually available, that can be used to uprate earnings as well as rental and other living expenses. In order to take into account that these variables would generally change differently for specific parts of the population, it is important to use uprating indices on the most disaggregated basis possible. In other words, where available, different uprating indices should be applied to men s and women s earnings, rental expenses in urban/rural areas, etc. An uprated monetary variable implicitly assumes that the structure of the underlying population has remained unchanged. In many cases this will not be realistic (Harding, 1996; Merz, 1991). The data may, for example, have been collected during a period of low unemployment, while the policy year we are interested in may be characterised by a much higher unemployment rate. If there are long gaps between data collection and simulation reference year, many demographic dimensions may change as well (ageprofiles, fertility, female labour force participation, etc.). There are basically two techniques for taking these changes into account: static and dynamic. Here we only discuss the static technique as dynamic adjustments are beyond the scope of both this report and the proposed model building project. The static ageing approach requires adjustment of weights in the data without actually altering the variables of any observations. Instead, the weights (e.g., those of employed and unemployed people) of 35

39 observations are adjusted to meet certain criteria (e.g., the level of unemployment). By specifying conditions that have to be met by the re-weighted dataset ( restrictions ), such as a set of aggregates and/or distributions that the data should reproduce, it is possible to find a new version of weights, which approximate the specified conditions. However, no firm statements can be made about how representative the new data are about dimensions that have not been controlled for (particularly, if there is little correlation with the control variables). To partially address this problem, some techniques have been developed to re-weight in a conservative manner. To address these issues we need to collect information and use it in as disaggregated a manner as possible. Variable uprating from the data year to the policy year can be difficult and is not always possible. Therefore it is essential sometimes to make a qualitative adjustment using historical sources. The indices should be for different measures of earnings growth by gender, region, industry, etc. The differences in the structure of a population such as unemployment participation rates, migration and so on have to be taken into account. Therefore it may be necessary to create an alignment or calibration factor to adjust the model to represent the policy year Tax Evasion This is the most controversial area of investigation and difficult for building a model. Detailed knowledge of the tax rates in circumstances without tax-evasion is essential in order to specify the correct financial trade-offs of individuals when choosing how much income to report to tax authorities. For empirical estimations these tax-benefit models can provide an input to the econometric specifications of tax-evasion models and detailed taxrate measures, as explanatory variables are taking into account the type of income, family situation, benefit payments, etc. It is obvious that in order to empirically estimate a model of tax-evasion, we require data describing the amount of evaded taxes for the population of interest. On a micro-level, these data are impossible to obtain. The estimation of tax invasion in Lithuania should not be a big problem. Normally medium and large size enterprises are quite good in declaring tax to the authorities. However, problems could arise with the self- employed, farmers, and small-size companies because Lithuanian society has not tended to reveal their true income level due to the high income tax rate. One possibility in addressing this issue is to use tax payment data that are available on the HBS and compare them with the simulated tax payment together with other records available from the Office of National Statistics. 36

40 6.1.6 Non- Take Up of Benefits Rapidly changing socio-economic contexts put increasing pressures on systems of meanstested benefits. Therefore it is important for both policy-makers and researchers to understand the principles underlying these types of benefit as well as the drawbacks associated with them. In recent years the phenomenon of non-take-up has arisen, where people who are eligible for a certain means-tested transfer payment do not apply for it. While non-take-up represents an interesting economic puzzle on a theoretical level, it also has important policy implications. It renders last-resort safety nets less effective in terms of ensuring a socially acceptable minimum of financial resources. More generally, a programme that does not reach its designated recipients will fail to fulfil the purpose it was designed for. From a policy evaluation point of view, taking into account non-takeup is essential to uncover the likely effects of policy reforms in terms of poverty reduction, income distribution, incentive effects and costs. This is especially true for the application of micro-simulation models and, more generally, all studies looking at the effects on low-income groups who frequently are the primary target of policy reform proposals. Similar to our suggested approach to tax evasion, the micro-simulation model can be used to simulate theoretical eligibility and theoretical benefit amounts for a representative sample of each country s population. This approach permits the calculation of theoretically available benefit amounts, for each potential benefit recipient, taking into account all relevant interactions between the various instruments of the tax-benefit system (income tax, social insurance contributions, universal and means-tested benefits). By comparing these simulated values with those actually reported by the individuals in the representative sample, one can then obtain an estimate for (a) the fraction of eligible people who do not claim the benefit (the non-take-up rate), and (b) the fraction of available benefits that are not claimed. Comparing the difference between actual and theoretical benefit amount, is to study the determinants of take-up behaviour. The information on the determinants of take-up behaviour would enable model builders to fine-tune their models and provide more realistic evaluations of existing social and fiscal policies as well as policy reforms. The tax-benefit model can assume 100% take-up and use the micro-simulation model to compute the potential effects of means-tested benefits in a situation where everybody applies for all transfer payments to which they are entitled by law. These results could then be compared to a second scenario where the model would be constrained to assign transfer payments only to those people who in fact receive them. It is possible to create a specific function in the model that deals with nontake-up. The difference between the two scenarios will give a useful indication of the importance of non-take-up. It illustrates the extent to which the effectiveness of existing means-tested benefits can be improved by finding ways to get financial resources to their designated recipients without requirement to alter the formal structure of transfer programmes. 37

41 Previous research has identified numerous factors potentially influencing take-up behaviour (Moffit, 1980; Moffitt, 1983; Cowell, 1986; Anderson and Meyer, 1997). Amongst them are: Factors associated with the stigma of being a benefit recipient (urban/rural area, regional unemployment, etc.); Factors associated with the cost of obtaining the benefit (cost of obtaining relevant information, strictness of rules, anonymous form or personal interaction with benefit agency, burden of proving eligibility, etc.); Factors associated with the potential gain of receiving the benefit (benefit amount, likely duration of low-income situation, degree to which benefit has to be paid back once self-sufficiency has been restored, etc.). Factors specific to Lithuanian circumstances affecting non-take ups of an eligible benefit are slightly different from those mentioned above. There have not been any particular studies in this area to underline the extent of non take-up of benefits. This is due to the lack of interest by Local Authorities providing support to the local people in need. The local government bodies hold a negative view of people in need. Local Government receives money from the Government budget revenue in order to pay benefits. If the money is not spent by the Local Authorities on benefits then it could be used for other purposes. This has resulted in an increased negligence of people in need and it is simultaneously accompanied by a growth in their number. However this system would be abandoned in 2006 and money that is not spent on support would have to be returned to the Government revenue. This would help estimate entitlement and take-up of benefits. At this stage it is difficult to define the characteristics of people who decide to either take up their entitled benefits or not. It is still possible to estimate 100% tax benefit entitlement and compare it with Local authorities expenditure on social benefits. Unfortunately it would be impossible to do it on the micro-level although on aggregate level it could give a spatial take-up rate for different benefits. 7. MAINTENANCE, SUPPORT AND TRAINING The tax benefit model-building project needs to take into account not only the construction of the model itself and related documents but also model maintenance. It can only be achieved by continuing co-operation between model builders and users. Due to the complexity of these models, questions and additional demands as well as possible errors in the model are only revealed via actual use of it. Close working relationships, continuing support for users and their feedback are essential. In addition, tax-benefit models are only useful for contemporary policy analysis if they are constantly updated. Policy rules built into the model must reflect the current circumstances. Accordingly, new 38

42 or improved micro datasets must be incorporated into the model as they become available and improve the quality of the model and also add additional functionality. In fact, most of these tasks can and should be done by qualified (trained) model users themselves, but there can be a degree of uncertainty of action points that will invariably require some input from model builders. Users of the model will need to be trained in both model use and maintenance. Therefore providing essential training (user manuals, technical guides and other relevant documents) is important. The most efficient way of user training and exchanging experiences is by providing face-to-face introduction courses, when practical use of the model can address useful policy analysis and questions. These exercises can involve both hypothetical case-studies and real-world evaluation scenarios. To ensure user confidence in modelling, additional courses on updating the model can be organized. User manuals and all other relevant information have to be provided. It is essential to put information in a clear way and simple language for potential users from different backgrounds. It would be useful to present it to potential users (Lithuanian Government Departments). In addition, we would propose to hold workshops with no more than four to six participants each and lasting three to four days to uncover the basics of model operation and to practice the use of the model by analysing case-studies and real-world policy questions prepared by the model builders in conjunction with potential users and other experts from Lithuania. The project s founders and future model users need to establish pragmatic and workable solutions to ensure the required level of ongoing user support. This is essential particularly in the first few months after delivery of the model. Establishing a Frequently Asked Questions database would enable users to find answers to their questions. 8. FINAL REMARKS As the nature of this work is a feasibility study, at present it would be difficult to predict the implementation stage. We do not have any information available about numbers of people being involved in building the model. Therefore we would not create any preliminary work plan for the project. Later it would be possible to take this project forward depending on the recourses available. It may change depending on new information about the Lithuanian tax-benefit system, available data or institutions and people involved in a model building project; this also has to include training and continuing user support. 39

43 9. SUMMARY OF TAX-BENEFIT SYSTEM RECOMMENDATIONS Table 8: Lithuanian tax-benefit system and recommendations for simulations Instrument Income tax Can be Simulated (yes/partly /no) 1 Yes for Majority (excluding disabled) Variable requirements TAXES Gross earnings Marital status Number of dependents Additional information needed Status of Disability information for household members VAT Yes Consumption of different goods and services Price of different goods and services Excise taxes Yes Consumption of goods and services that are taxed Price of these goods and services Dividends Data Income variable from dividends Taxation of inherited property Data Income variable from inheritance Property tax Data Property tax variable Land taxes Data Land tax amount Employees social insurance contribution Employers social insurance contribution Self-employed social insurance contribution Yes Gross earnings (transformation Net incomes to Gross) Yes Gross earnings (transformation Net incomes to Gross) Partially Income from selfemployment Type of selfemployed 40

44 Contributory benefits and Social Benefits Old-Age Pensions Data Pension variables (amounts) data on employment history Disability pension top-up Data Pension variable data on employment history and disability status Survivor Benefits Data Benefit data amount Information on dead family member Sickness or temporary disability benefits Data Disability benefit variables (amounts) Information on previous salary Maternity benefit Data Information on child birth Information on last salary of the mother Information on last insured salary Maternity benefit variables (amounts) Workers Medical no Benefits Unemployment benefit Data Information on receipt of this benefit Educational attainment Permanent incapacity Data Amount of a benefit Disability status Funeral benefit Data No data on deaths in HH information on family member s decease Birth grant yes Information on child birth Number of children Child birth benefit Yes Information on child birth Number of children Maternity (Pregnancy) Data Amount of a benefit data of pregnant Benefit for Studying students is not Women available Benefit for Children of Data Amount of a benefit Military Conscripts Orphan s stipend Data Amount of a benefit Information about parents Foster Child Benefit yes Variable on relationship Information about parents State support income Yes Income variables Family size Benefit for temporary incapacity of work Data amount of this benefit Information on last 6 months salary 41

45 Death Grant Data Amount of this benefit Social pensions Data amount of this benefit Age information on family member s decease Degree of disability Information on previous employment Care for totally disabled person Data Amount of this benefit Degree of disability Compensation of Data Benefit amount transportation expenditures Housing Data Benefit amount Local authorities housing rates, and data needs to be representative on a LA level Notes: 1 the entry (data) means that the amount for this instrument will be taken from the data (rather than simulated). 42

46 REFERENCES Alphametrics, 2002, Analytical and statistical tools for monitoring EU tax-benefits systems, Unpublished report. Anderson P M and B D Meyer, 1997, Unemployment Insurance Take-up Rates and the After-Tax Value of Benefits, Quarterly Journal of Economics, 112 (3), pp Atkinson A, F Bourguignon, C O Donoghue, H Sutherland and F Utili, 1999, Microsimulation and the Formulation of Policy: A Case Study of Targeting in the European Union, EUROMOD Working Paper EM2/99. Atkinson A B, 2002, "Evaluation of National Action Plans on Social Inclusion: The role of EUROMOD", EUROMOD Working Paper EM1/02, available: Atkinson A B and F Bourguignon, 1990, Tax-benefit Models for Developing Countries: Lessons from Developed Countries, DELTA Working Paper Bernotas, Dainius, Guogis, Arvydas, Evaluation of Social Security in Lithuania, Latvia and Estonia: Achievements and Drawbacks, COST A 15 project Reforming Social Security Systems in Europe (No.V-057). Caldwell, S., and Morrison, R.J., 2000, Validation of longitudinal dynamic microsimulation models: experience with CORSIM and DYNACAN, in: L. Mitton, H. Sutherland, M. Weeks (eds.), Microsimulation Modelling for Policy Analysis. Cambridge University Press Clotfelter C T, 1983, Tax Evasion and Tax Rates: An Analysis of Individual Returns, Review of Economics and Statistics, 65 (3), pp Cowell F A, 1986, Welfare Benefits and the Economics of Take-Up, Discussion Paper No 89. Creedy J and A Duncan, 2002, Behavioural Micro-simulation with Labour Supply Responses, Journal of Economic Surveys, 6 (1), pp Gylys, Povilas, 2004, Reforms of Pension System in Lithuania, Harding A, 1996, Introduction and Overview, Micro-simulation and Public Policy, Amsterdam: North Holland. Immervoll H, 2000, The Impact of Inflation on Income Tax and Social Insurance Contributions in Europe, EUROMOD Working Paper EM2/00. Immervoll H and C O'Donoghue, 2001a, "Imputation of Gross Amounts from Net Incomes in Household Surveys: An Application using EUROMOD", EUROMOD Working Paper EM1/01. Immervoll H and C O'Donoghue, 2001b, "Towards a Multi-Purpose Framework for Tax- Benefit Microsimulation. EUROMOD Working Paper EM2/01. Immervoll, Herwig, Tarcali, Géza, 2002, Development of a Tax-Benefit Microsimulation Model for Moldova, Unpublished Report. Lazutka, R. Tikrasis pensijø reformø tikslas valstybinë protekcija verslo rûðiai // Martini, Alberto, Trivellato, Ugo, "The Role of Survey Data in Microsimulation Models for Social Policy Analysis", Labour, 11, , April 1 43

47 Merz, J., Microsimulation- a survey of principles, developments and applications", International Journal of Forecasting, 7, , May 1 MINISTRY OF SOCIAL SECURITY AND LABOUR ( LIETUVOS RESPUBLIKOS, 2002 m. liepos 2 d. Nr. IX-1007, GYVENTOJŲ PAJAMŲ MOKESČIO ĮSTATYMAS Moffit, Robert, 1984, "Profiles of Fertility, Labour Supply and Wages of Married Women: A Complete Life-Cycle Model", Review of Economic Studies, 51, , April 2 O Donoghue, Cathal, "XLMSM DOCUMENTATION" (work in progress) Pechman, Joseph A., and Benjamin A. Okner, "Who Bears the Tax Burden?" Washington, D.C.: The Brookings Institution 44

48 APPENDIX 1: Data from Lithuanian Government Departments Figure 4: Money flow within the pension system of Lithuania 45

49 Figure 5: Family Benefit Formulas Source: Presentation Tax-benefit System in Lithuania by Romas Lazutka,