Tax revenue and labour income taxation in France
Tax revenue as a share of GDP is a usual measure aiming at comparing different tax-and-benefit system in time and across country. Figure 2 presents the evolution of the composition of tax revenue in France since 1949. If indirect taxation was the main source of tax revenue in 1949 (50%), its contribution decreased up to 25% of tax revenue in 2016. This was offset by an increase in the share of social security contributions and payroll taxes that grew from 30% in 1949 to more than 40% in 2016. The third element by importance in terms of tax revenue is the (labour and capital) income taxes, which amount today to about 20% of tax revenue. Yet, the creation of a flat income tax in 1991 fundamentally modified the structure of taxation by introducing a new way to fund social protection in France.
There are three main types of taxes on earnings in France in 2018. The first one are the social security contributions (SSCs), divided into employer and employee SSCs. In that cat-egory I also include other payroll taxes that are not affected to the funding of the social security because they share the same tax base2 . On figure 2, they correspond to the “SSCs and payroll taxes” category.
Second, wages are subject to the standard progressive income tax on a tax base less inclu-sive than the gross wage. Indeed, fiscal households can deduce business expenses (itemized or flat-rate allowance) as well as other deductions to the income tax base. Third, a flat in-come tax (contribution sociale généralisée, CSG) was introduced in 1991. The coexistence of a both a flat and a progressive income tax had been the norm ever since the creation of the income tax in 1914 and until 1959 when the flat part of the schedule was dropped. Thus, with the CSG, a flat tax is reintroduced but it is much less salient than the income tax is, as it is directly levied from the payroll), and does not suffer from the many tax loopholes of the progressive income tax. As a result, the CSG tax revenue progressively increased and is today almost equal to the income tax revenue.
2 Taxes on payroll correspond to the national account category taxes sur les salaires et la main d’œuvre. For example, the apprenticeship taxe, the tax on payroll, the contribution to the National Fund for Housing Assis-tance (FNAL) and the versement transport are taxes on payroll.
Social security since 1945: creation and evolutions
Social security contributions constitute the major component of labour income taxation and have undergone large evolutions. The social security system is characterized by the exis-tence of several regimes and many different types of contributions. Workers contribute to the regime that is specific to the sector of their firm. When the main regime was created in 1945, it was intended to incorporate all workers in a unified framework. When extended to the farming sector and to the self-employed it did not result in a unique and compre-hensive regime but rather in a constellation of different regimes. Overall, the tax schedule depends on tax rates applied to tax brackets whose thresholds are multiple of the “social se-curity threshold” (SST). The tax base is the gross wage (or posted wage). The contributive part of SSCs gives rights to incomes differed in time, such as unemployment and old age pensions, of which amount depends on the amount contributed. The non contributive part of SSCs funds benefits that do not depend on the amount contributed, such as family and health contributions.
Since the 1970s, wage taxation in France underwent a series of reforms that increased the progressivity of SSCs and decreased the weight of the progressive income taxes. First, employer social security rates, which applied under the part of the gross wage inferior to the SST, were progressively applied to the totality of the wage. Second, some non-contributive employee social security contributions were transformed into a new flat income tax, whose tax base is larger than wages and encompasses labour and capital income. This constituted a first step toward changing the paradigm of 1945 on which stands social security. Indeed, social security from then was not to be financed anymore only by labour incomes. Third, starting in 1993, a series of reductions of employer SSCs applied on low incomes, simultane-ously with an increase of minimum wage. These reforms increased the net wage at the level of the minimum wage without increasing the labour cost. They also had a redistributive pur-pose by increasing the progressivity of the SSCs schedule. The reductions were financed by public spending, decreasing again the share of social security spending financed by labour income.
At the origin, the French social security system was designed according to the principles of the bismarckian model of social protection system. Yet, the system relies more and more on fundings from the income taxes, which gives it a beveridgian component as well. This dual affiliation to canonical models of social insurance attests the peculiarity of the French social security system.
Whereas income inequality increased during the last thirty years in the United States, it stayed relatively stable in Europe (Alvaredo et al., 2018). This differential evolution questions the determinants of inequality. Many other factors can affect inequality trends, among which technology or globalisation. After documenting recent evolutions of inequality, I present the literature aiming to explain these evolutions. I then turn to questions more specifically linked to taxation.
Documenting the increase in inequality
Starting with Kuznets and Jenks (1953), a strand of literature aiming at describing the long term evolution of income inequality emerged. The field experienced a rebirth with Piketty (2001) and rapidly grew up to reaching a new step with the WID.world project (World Wealth and Income Database) and its companion website (http://wid.world) with time series on pre-tax income shares by country. This literature is based on a main inequality measure, the share of income accruing to a specific group among the total amount of income. The under-lying income concepts, the pre-tax income, is both consistent with national income from the national accounts and with the income distribution as observed in fiscal documents.
Evolution of income inequality
In a survey of the literature on the long term evolution of inequality in the 20th century, Atkinson et al. (2011) compare the evolution of top income shares across countries. They show that most countries experience a common decrease in income inequality during the first half of the century due to major historical events such as the two world wars and to the depression of 1929.
The dynamics evolution of income inequality haved diverged across countries starting in the 1960s. It increased in English-speaking countries (Piketty and Saez, 2003) but remained roughly stable in European countries (excluding the United Kingdom). In France, Piketty (2001) documents the evolution and dynamics of income inequality, showing how the pro-gressivity of the French fiscal system prevented the rise in capital income inequality. Garbinti et al. (2017) build on this previous work, improving the methodology, and show that income growth has been three times larger in the very top percentiles of the income distribution (3%) than in the rest of the income distribution since 1983.
The underlying methodology relies on the use of aggregate and individual tax returns data and national accounts. It consists in computing time series of income shares accruing to certain groups of individuals, based on incomes consistent with national accounts’ concepts.
Inequality at the top
The top of the income distribution has received substantial academic attention. Of course, the fact that top earners bring a large share of tax revenue justifies this interest. This renewed interest is also explained by the fact that top income shares started to increase again in the United States in the 1980s, as underlined by Alvaredo et al. (2013). Top income shares are the main measures of the evolution of inequality.
The relative importance of capital or labour income contribution to inequality evolved over time. If capital income was overwhelmingly important in explaining total inequality at the beginning of the 20th century, the capital destruction caused by the two World Wars and the high level of taxation undermined the importance of capital in the total income of the richest for a long time (Piketty and Saez, 2003). Yet, this trend seems to have turned (Piketty, 2013), with a coming back from capital income.
In the meantime, labour income contributed significantly to the recent increase in in-equality. Piketty and Saez (2003) show that wages played a major role in the increase of the top 1% share observed since the 1980s. They emphasize the role of what they call the “work-ing rich”3, a population of individuals earning such high wages that they reach the top of the wealth distribution with no other sources of incomes. Godechot (2007) (updated by Gode-chot (2017)) studies an example of a “working rich” population, the financial industry. He provides both an extensive fieldwork and a quantitative analysis on the wage setting and bonus attribution processes. Growing availability of administrative data allows him to study more precisely the factors at play in the increase in income inequality. For example, he uses payroll tax data to study the role of the financial sector in the evolution of wage inequality (Godechot, 2012), showing that the financial sector contributed to 48% and to 57% of the rise of the top 0.1% and top 0.01% income shares between 1996 and 2007.
Explaining the increase in wage inequality
The increase in wage inequality observed in many OECD countries since the seventies gave rise to a strand of the labour economic literature dedicated its causes. I decompose the ex-planations according to the part of the labour income distribution they target. I present first the theories addressing the increase in inequality of the whole income distribution. I focus then on the arguments specifically addressing the inequality increase at the top of the labour income distribution.
Inequality among the bottom 99%
Two main families of explanations for the increase of income inequality on the labour market are traditionally opposed.
The first set of theories considers that the demand-side of the labour market is the main driver of the increase in inequality. The central factor at play is technological change. The skill-biased technological change assumption, which has it that technology affects the or-ganisation of labour and increases the need for skilled labour relative to unskilled labour, constitutes a first interpretation of this explanation. This assumption has now been tested several times (Katz and Murphy, 1992; Card and Lemieux, 2001) and for several countries: Autor et al. (2008) for the United States, Dustmann et al. (2009) for Germany, Lindley and Machin (2011) for the United Kingdom. Recently, another explanation emerged from the same trend: technological changes would cause a job polarization (Autor et al., 2006; Hunt and Nunn, 2017), with fewer jobs in the middle of the distribution relative to the low-wage low-skilled and high-wage high skilled jobs.
A second strand of explanations has it that institutional factors such as the minimum wage (Card and DiNardo, 2002) and unions (Fortin and Lemieux, 1997) played an important role in the increase in inequality. With Autor et al. (2008), a consensus toward demand-side explanations emerged in the literature.
Focus on the role of top wage earners
If the previous factors influence the evolution of the whole labour income inequality, top earners recently drove the increase in inequality. The explanations for the wage increase of top earners can also be classified into two main categories, opposing market-based and institutional-based arguments.
A first set of explanations develop arguments relying on the determinants of the supply-and-demand for top wage earners. Gabaix and Landier (2008) proposes a model of the com-petitive labour market of CEOs where the firm size explains the large differences in pay. In that context, the increase in firm sizes caused by increases in market capitalization would explain the increase in CEOs’ pays. The same technological argument than before applies to top wage earners. Kaplan and Rauh (2013) provide evidence in line with the skill-biased technological change, showing that higher returns to skills were particularly important for top wage earners.
A second set of arguments highlight the role of institutional factors. According to Bertrand and Mullainathan (2001), firm’s governance plays an important role as CEOs would be more able to capture profit unrelated to their actions in poorly governed firms. The level of tax-ation can also be considered as an important institutional factor affecting the wage setting process of top wage earners. Piketty et al. (2014) show that lower top tax rates increase the bargaining power of top wage earners.
The rapidly expanding documentation of the evolution of inequality brought back on the re-search agenda the controversy on the economic factors affecting labour income distribution. Yet, one classic explanation in the public finance literature is too often set aside by the labour economics literature. Indeed, taxation on the labour market is a potential driving force of the evolution of wage inequality. The next section is dedicated to presenting the questions and problematics from the taxation literature relevant for the study of the wage distribution.
The impact of wage taxation
The extent to which taxes and benefits policies succeed in reducing post-tax income inequal-ity is an empirical question. The direct effect of a policy might be counteracted by its impact of the behaviour of agents (individual and firms for example). Moreover, some political tools might be more effective than others for attaining a same level of redistribution. As an illustra-tion, Doerrenberg and Peichl (2014) show using a cross-country analysis that social expen-diture policies are more effective at reducing inequality than progressive taxation because of behavioural responses. Evaluating the impact of wage taxation relies on complex models and numerous assumptions. I first expose a methodology used for assessing the redistributive impact of taxes and benefits. I then present the findings of the literature regarding the two main assumptions underlying the previous methodology, tax incidence and the behavioural responses to taxes.
The redistributive impact of taxes and benefits
Assessing the redistributive impacts of taxes requires first to describe the taxes paid depend-ing on the income amount, the income type and on some demographic characteristics. This is commonly achieved by micro-simulation models.
Invented by Orcutt (1957), micro-simulation is a method aiming to empirically look at the theoretical impact of change on a given population. Typically, micro-simulation meth-ods are used to study ex-ante the impact of a tax change on the distribution of individual income. Models are static when there is only one period and dynamic when encompassing several time periods. First order effects of policy changes are given by their direct impact on the individual disposable income. For that purpose, every model relies on incidence as-sumptions regarding the range of taxes and benefits implemented. Yet, reforms can also have second order effect when triggering changes in behaviour. Behavioural modules relying on behavioural assumptions can be incorporated to micro-simulation models in order to en-compass these side effects. The underlying assumptions ruling the tax incidence and the behavioural responses are of major importance. They must be clearly stated and discussed.
Another important feature of a micro-simulation model is whether or not it is consistent with national accounts. Such a consistency ensures that incomes and taxes are in line with the amounts from the national accounts. This is useful for estimating the closest aggregate effect of a reform.
The incidence of wage taxation
The theoretical literature on incidence shows that even if a tax is legally borne by individuals or households, the economic incidence differs from the legal one. Models solve the puzzle of the economic incidence in theory (see Fullerton and Metcalf (2002) for a review of the theoretical models in partial and general equilibrium) but there is no empirical consensus on this important question of public economics.
The question is particularly relevant regarding wage taxation. Indeed, payroll taxation is legally paid by employers for employer social security contributions and by employees for employee social security contributions. Yet, the economic incidence is not necessary equal to the nominal or legal incidence. The theoretical incidence depends on the elasticity of sup-ply and demand for labour. Empirical evidence on the incidence of taxation on labour supply and wages is still scarce and there is no consensus. The recent literature uses micro data and exploits social security reforms. But evidence goes from full-incidence on employers (Saez, Matsaganis and Tsakloglou, 2012) to full-incidence on employees (Gruber, 1997). New con-tributions are made by focusing on the working context or on the specificities of the taxes. For example, Saez et al. (2017) shed light on the importance of the firm environnement. They rely on the introduction of an age-dependent payroll tax rate cut and show that the incidence was fully on employer but that the reform also had employment effect and firm-level effects. Fur-ther, the incidence of difference taxes on a same income might differ, as shown by Lehmann et al. (2013). The tax-and-benefit linkage seems also to matter (Iturbe-Ormaetxe, 2015; Bozio et al., 2017).
Estimating behavioural responses to taxation
Behavioural responses are inherent to taxation and should be taken into account both in theoretical models deriving optimal tax schedules and when designing public policies. They can be decomposed into three main categories. The labour supply is a first channel of real economic response encompassing changes in hours worked, migration (Kleven et al., 2013, 2014) and work effort. Optimisation behaviours go through time-shifting of income (Gools-bee, 2000; Kreiner et al., 2016), income-shifting (Pirttilä and Selin, 2011; Harju and Matikka, 2016) or the use of deductions. A last possible response consists in evading the tax, which is not legal, as opposed to optimisation.
Rather than studying each response individually, the new modern public finance litera-ture focuses on a central parameter, the elasticity of taxable income (ETI) with respect to the net-of-tax rate4. The ETI encompasses the total behavioural responses to tax rates. Since the seminal contributions of Feldstein (1999), this strand of literature has evolved. The recent developments were surveyed by Saez, Slemrod and Giertz (2012) in a comprehensive litera-ture review. This section reviews some aspects of the recent literature with a specific focus on top income earners.
Sufficient statistic One of the motivations for estimating the ETI is that this parameter would be a sufficient statistics for welfare analysis under some assumption5. Feldstein (1999) proposes a theoretical framework where estimating only the ETI parameter is sufficient for estimating the dead-weight loss of income taxes.
The sufficient statistic concept is at the heart of the methodological debate opposing structural versus reduced-form methods in policy evaluation. The first approach focuses on the estimation of primitives of a theoretical model and uses the result to simulate the im-pact of policies on welfare. The second approach relies on exogenous sources of variation for identifying causality. While structural approaches often require very strong assumption, reduced-forms approach are subject to critique à la Lucas since estimates are doomed to depend on the policy context.
The interest for the sufficient statistic approaches in public economics came back with Chetty (2009). This article contributes to bridging the gap between the two paradigms by showing how the applied economics literature has made the synthesis between the two meth-ods. Recent developments in this literature provide extension of the theoretical models such as Kroft et al. (2017) and highlight some caveats. For example, Doerrenberg et al. (2017) show that for tax systems with deductions, the ETI might not be a sufficient statistics if the tax deductions are assumed to generate externalities and if deductions are responsive to tax changes.
Data and estimation strategies The ETI generally is estimated by difference-in-difference methods relying on three different types of data. The first evidence relied on time series of income shares. The increasing availability of micro files of income tax returns led to cross-sectional and panel data analyses.
Two main identification problems arise. The first one comes from the endogeneity of the tax rate. Auten and Carroll (1999) address this issue and propose an instrumentation strategy based on the predicted net-of-tax rate based on previous year income. A second issue arises with strategy based on panel data, the mean-reversion issue, and is particularly problematic for the richest individuals. Weber (2014) compares the existing strategies and proposes a new method for panel data addressing both identification threats.
Results Thus far, ETI estimates range between 0.1 to 0.8 and are heterogeneous in the pop-ulations. Self-employed are among the more responsive. Top income earners also demon-strate larger responses, optimisation and avoidance.
This section gave a broad overview of the questions, the findings and the methods I rely on, mixing imports from the public finance and the labour economics fiewlds.
The previous section gave a broad overview of the literature on which the thesis stands. The aim of this section is to state how this work fits in and contributes to this literature. I show how I connect the labour economics and the public finance literature, relying on the use of administrative data.
Pre-tax versus post-tax earnings concepts
The labour economics literature develops arguments tackling the increase in pre-tax in-equality. Indeed, the set of explanations emphasizing the role of the demand-side of the labour market implicitly relies on a comprehensive wage concept. For example, the theo-retical framework developed to test empirically for the existence of a skill-biased technolog-ical change relies on the equalization of workers’ wage and marginal productivity. In the real world, the closest concept to the theoretical wage is the labour cost wage. Even if it is not made explicit in the literature, the concept of labour cost applies well in theoretical models. Yet, labour costs are more often then not unavailable when it comes to testing the models, and the empirical analyses rely on the data which is available. Most of the time, administra-tive data contain information on posted wages only whereas survey data inform about net wages.
This is precisely where insights from the public finance literature can help filling this empirical gap. Indeed, this literature focuses on the wedge introduced by taxes between the pre-tax and the post-tax income. The pre-tax and post-tax concepts are constitutive of the fields and of the methodologies developed, such as microsimulation. By using the mi-crosimulation methodology from the public finance literature I contribute to the strand of the labour economics literature focused on explaining market wage inequality by simulating labour cost. Disentangling between the net wage and the labour cost in the empirical anal-ysis does not only consitute a methodological argument. It also introduces a fundamental difference between the concept of earnings relevant from the employer’s and from the em-ployee’s point of view. The introduction of a discrepancy between the two wage concepts is new to the labour economics literature, that focused on a simpler and unique market wage.
Another way of stating this point is to emphasize the role of taxation, which is at the root of the difference between the net wage and the labour cost. More precisely, the difference between the labour cost and the net wage comes in France from social security contribu-tions and from the flat income taxes6. Enriching the analysis with this dual wage concept has an impact if the labour cost and the net wage did not evolve in a parallel fashion. In most countries, their long-term evolutions did not differ, or at least did not differ according to the income level. Yet, France experienced large reforms of SSCs since the 1970s that increased their redistributive power. I contribute to the literature by documenting the impact of the evolution of the legislation on SSCs on the pre- and post-tax labour income inequality.
The argument in favour of the differentiation between the labour cost and the net wage is likely to have a long term impact on the labour economic field as it is relevant for any theoretical model of the labour market. Yet, the simulation of labour cost comes at a high fixed cost, which values all the more this contribution.
Overall earnings distribution versus top earners
The empirical public finance literature aims to assessing the redistributive power of the tax system, tackling the question of incidence and of behavioural responses. Top income earners are shown to be the most responsive to taxation. The literature mainly relies on the evolution of the income shares accruing to individuals at the top of the distribution, relating them to the marginal rate of taxation. The findings disentangle between several forms of behavioural responses and different types of incomes. Yet, little is known about the specific context in which these top earners evolve, except for very specific individuals (football players for ex-ample). The top of the income distribution is often opposed to the rest of the distribution. Yet, the threshold between these two categories is often ad hoc and the categories are es-sentialized. Because the administrative data do not contain non tax related information, the literature cannot detail characteristics of responsive population, such as the occupation, the educational attainment or the employment sector.
Conversely, the labour economics literature studies top labour income earners in their working environment. For example, a large strand of literature focuses on the determination of CEOs’ pay in the context of the firm. I use this approach where workers belong to the firm context in order to shed a new light on the taxation process of individuals at the very top of the earnings distribution. This allows me to enrich the classical public finance questions by encompassing the worker-firm relationship. Especially for top earners, firm outcomes and choices can be an important channel of response to taxes.
The use of administrative data has been driven by increasingly facilitated access to researchers. They provide new sources for the empirical assessment of the evolution of earnings and gave rise to the development of new methodologies.
The literature on inequality and taxation now relies on administrative tax data that are crucial for assessing correctly the level of inequality because they are not subject to the under-declaration of income as standard surveys are. They also cover accurately the top of the income distribution and provide details on the income type. The labour economics literature also saw an increase in the use of administrative data, with the use of matched employer-employee datasets. The main interest of these data is to inform on the whole em-ployment structure of firms. I take advantage of the fact that employer-employee datasets actually come from payroll tax returns: they are by definition tax return data.
Yet, these data are not without limitations. First, they rely on tax concepts of incomes that need be transformed into economic concepts, using microsimulation for example. I demon-strate that it is possible to simulate labour income taxes based on payroll tax data, applying the standard simulation method from public finance in the specific case of labour incomes. Second, only taxable incomes are reported, meaning that non taxed incomes are not con-sidered. For instance, incomes such as fringe benefits are not encompassed in this data. On top of that, these data suffer from the same shortcomings as does the income tax, such as tax evasion.
Cross-utilization of several administrative sources help with these issues by allowing re-searchers to identify some income responses. For example, using a panel of income tax data makes it possible to highlight avoidance along the time dimension. With this in mind, I de-velop an algorithm building an individual panel of the universe of payroll. Matching individ-ual and firm tax data also enables me to identify the individual-level impacts of firm-level taxation as well as the firm-level impacts of individual-level taxation. This constitutes an-other contribution that from labour economics into the taxation field. Indeed, it is now classical in labour economics to encompass both an individual-level and a firm-level dimension (with individual and firm fixed effects for example) but it seems that it is not enough the case in the tax responsiveness literature. Some seminal articles highlight potential substitution between business and individual income, but there is no compelling evidence bringing to-gether firm corporate tax data and payroll tax data. I provide such an example by studying both employee- and employer-level of response to the 75% tax on millionaires.
Table of contents :
1 Overview of labour taxation in France
1.1 Tax revenue and labour income taxation in France
1.2 Social security since 1945: creation and evolutions
2 Literature review
2.1 Documenting the increase in inequality
2.2 Explaining the increase in wage inequality
2.3 The impact of wage taxation
3 General problematic
3.1 Pre-tax versus post-tax earnings concepts
3.2 Overall earnings distribution versus top earners
3.3 Administrative data
4 Outline of the thesis
4.1 Fiscal policy and redistribution in France
4.2 Primary inequality and taxation on the labourmarket
I FISCAL POLICY AND REDISTRIBUTION IN FRANCE
1 L’analyse redistributive des prélèvements obligatoires par microsimulation : méthodologie,
usage et limites
1 Présentation du modèle TAXIPP
1.1 Constitution de la base de données source
1.2 La simulation du système socio-fiscal
2 Analyser l’incidence de l’ensemble des prélèvements obligatoires
2.1 La théorie de l’incidence fiscale
2.2 Les mesures empiriques de l’incidence fiscale
2.3 Hypothèses retenues par le modèle
3 Mesurer la distribution des taux d’effort
3.1 La distribution des taux moyens de prélèvements obligatoires en 2010
3.2 Étudier la fiscalité des hauts revenus
4 Une approche qui a pourtant des limites
4.1 Sensibilité aux hypothèses d’incidence
4.2 Conflits dans les sources de données agrégées
5 Conclusion et perspectives
2 French public finances through the financial crisis : a macro and a micro analysis
1 Impact of the financial crisis : the macro picture
1.1 National income
1.2 Labour markets
2 Public finance responses
2.1 Fiscal stance before the crisis
2.2 How did the crisis affect the public finances?
2.3 What was the fiscal response to the crisis ?
3 Policy responses : an opportunity for reform?
3.1 Changes to tax and benefits
3.2 Changes to public spending
3.3 Other structural reforms
II PRIMARY INEQUALITY, INCIDENCE AND BEHAVIOURAL RESPONSES ON THE LABOUR MARKET
3 Primary inequality of labour incomes: taxation and technological determinants
1 Labour cost inequalities in France
1.1 Labour cost and social security contributions
1.3 Labour cost inequalities in France (1976-2010)
2 Revisiting demand-side explanations of inequalities using tax changes
2.1 Supply, demand and wage premiumfor skilled workers
2.2 The standard supply/demand model with taxes
3 Impact of taxation on wage inequalities
3.1 Can we infer SSC incidence from SBTC?
3.2 Behavioral responses and the impact of taxes on wage inequality
4 Conclusive comments
3.A Institutional details on social security contributions and income tax in France
3.A.1 Uncapping of the social security contributions
3.A.2 Reduction of employer social security contributions
3.B Data andmethods
3.B.1 Data and variables
3.C Additional Figures
4 Who paid the 75% tax on millionaires? Optimisation of salary incomes and incidence in France
2.1 Taxing Labour Income Earners in France
2.2 The 2013-2014 reform
3 Data and descriptive statistics
3.1 Linked employer-employee dataset
3.2 Income tax return data
3.3 Simulation of income and payroll taxes
3.4 Descriptive statistics and graphical evidence
4 Worker-level and firm-level effects on wage
4.1 Wage incidence
4.2 Wages and employment at the firmlevel
5 Optimisation behaviour of the individual
5.1 Conceptual Framework
5.2 Cell-based approach
5.3 Repeated cross-section approach
4.A Panel regression approach
4.A.1 Building a panel data from administrative datasets
4.A.2 Estimation strategy
4.A.3 Estimation results
4.B Additional tables and figures