RAISING TAXES TO BALANCE THE BUDGET: HOW EFFECTS ON OUTPUT AND LABOR SUPPLY COMPLICATE THE MATTER

Transcription

1 1 RAISING TAXES TO BALANCE THE BUDGET: HOW EFFECTS ON OUTPUT AND LABOR SUPPLY COMPLICATE THE MATTER May 9, 2011 Gregory Hirshman Department of Economics Stanford University Stanford, CA Under the direction of Prof. Robert Hall ABSTRACT The United States and many major European countries currently are running large budget deficits, and most of these countries also confront significant national debt. Many are now seeking to restore fiscal discipline by cutting spending, raising taxes, or doing both. In this thesis, I explore what would occur if the United States or any of seven major European countries (France, Germany, Italy, the United Kingdom, Greece, Spain, and Sweden) attempted to balance its budget simply by raising taxes. I also determine what would occur if a country decided to balance its budget by using tax increases to accomplish half of its budget fix and cutting government spending to do the rest. To pursue this analysis, I first employ a simple static model to gain some understanding of how changes in tax rates affect how much people work. I then develop a more realistic and sophisticated dynamic general equilibrium model and gradually add refinements to it. With that model, I find that if government spending is not reduced, all countries except Sweden, which has only a small budget deficit, are either incapable of achieving a balanced budget in one year or could do so only by suffering dire economic consequences. Even if cuts in government spending account for 50% of the budget fix, all countries under analysis (except Greece which is unable to balance its budget under any of the four scenarios I posit) would experience at least some loss in output, and many would face considerable hardship. Keywords: Labor supply, tax rates, government spending, budget deficit, United States, Continental Europe, Scandinavia Acknowledgements: I thank Professor Robert Hall for overall guidance. I thank Professor Geoffrey Rothwell for suggestions on my study design. I also thank my parents who helped me edit this thesis and my brother Brian who helped me develop my MATLAB code.

2 2 Table of Contents Chapter 1: Introduction... 6 Chapter 2: Literature Review Labor Supply Across Countries The Effect of Tax Rates The Effect of the Way that Government Spends its Revenue The Effect of Labor Unions, Labor Market Regulations, and Productivity The Effect of Income Inequality Background of the Dynamic General Equilibrium Model The Neoclassical Starting Point Dynamic Modeling Chapter 3: A Static Model of Tax Rates and Labor Supply Background Theoretical Framework My Key Equilibrium Relation Explanation of Results Chapter 4: My Dynamic General Equilibrium Model... 45

3 How the Dynamic Model Improves upon the Simple Static Model Introduction to the Dynamic General Equilibrium Model Chapter 5: Employing My Dynamic Model The Trivial Case: No Taxes The Model with Taxes Changing the Tax Rate in the Dynamic General Equilibrium Model Increasing the Tax Rate from 20% to 30% Decreasing the Tax Rate from 20% to 10% Varying the Magnitude of the Change in the Tax Rate Removing the Consumption-Work Complementarity Assuming a Lower Wage Elasticity of Labor Supply Chapter 6: Alternative Scenarios for Government Spending Government Spending on Wasteful Projects The Concept of μ Results from Introducing μ into the Model Increasing the Tax Rate from 20% to 30% Decreasing the Tax Rate from 20% to 10%... 75

4 Implications of μ Chapter 7: Welfare Analysis Welfare Analysis over a Lifetime in the Baseline Case Utility Changes Caused by a Change in the Tax Rate Consumption Gain and Loss Equivalents Caused by Changes in the Tax Rate Hours Worked Gain and Loss Equivalents Caused by Changes in the Tax Rate Welfare Analysis Between the Steady States in the Baseline Case Utility Changes Caused by a Change in the Tax Rate Consumption Gain and Loss Equivalents Caused by Changes in the Tax Rate Hours Worked Gain and Loss Equivalents Caused by Changes in the Tax Rate Welfare Analysis Assuming No Complementarity or a Less Elastic Labor Supply Welfare Analysis Assuming Different Values of μ Chapter 8: Applying My Model to Real World Scenarios The Economic Situation in the Eight Countries Under Consideration Scenarios for Balancing the Budget in One Year Assuming Lump Sum Transfers with No Change in Spending Assuming Lump Sum Transfers with Spending Cuts 50% of the Budget Fix... 96

5 Assuming μ = 0.50 with No Change in Spending Assuming μ = 0.50 with Spending Cuts 50% of the Budget Fix The Consequences of Using Tax Increases to Balance the Budget A Closer Look at the Case of the United States Laffer Curves for the United States Balancing the Budget in the Long Run Chapter 9: Conclusion References Appendix A: Matlab Code Appendix B: Results Spreadsheets

6 6 Chapter 1: Introduction The United States and many major European countries are currently running large budget deficits, and most of these countries also confront significant national debt. As a consequence, many are now trying to restore fiscal discipline after providing generous entitlements, state pensions, and other government benefits for many years. Although tax rates have increased in recent decades in most advanced industrial countries, the resulting increase in government revenues has been greatly outpaced by the increase in government expenditures. Many countries must now cut spending, raise taxes, or do both. In this thesis, I intend to explore what would occur if the United States or any of the seven major European countries (France, Germany, Italy, the United Kingdom, Greece, Spain, and Sweden) attempted to balance its budget simply by raising its taxes. I also seek to determine what would occur if each of those countries decided to balance its budgets by using tax increases to accomplish half of the budget fix and by cutting government spending to do the rest. Can these countries succeed in balancing their budgets? If so, what would be the economic costs in terms of aggregate output and labor supply? To pursue this investigation, I will first develop a simple static model to gain some understanding of how changes in tax rates affect how much people work. I will then create a more realistic and sophisticated dynamic general equilibrium model and gradually refine it. With that model, I will return to my principle research questions. As historical background for this thesis, it is useful to understand the changes in labor supply that have occurred in the United States, Continental Europe, and Scandinavia. In the early 1970s, hours worked in the market sector by persons between the ages of 15 and 64 was approximately equal in the United States, Continental Europe (which I define as France,

7 7 Germany, and Italy), and Scandinavia (which I define as Denmark, Finland, Norway, and Sweden) (Prescott 2004). By the mid-1990s, however, Continental Europeans worked only about 70% as much as Americans, and Scandinavians worked about 85% as much as Americans (Prescott 2004). In the past decade, the proportional labor supply differences between both Continental Europe and the United States and between Scandinavia and the United States have remained relatively constant (Rogerson 2006). What caused the large decrease in the relative labor supply in Continental Europe between the early 1970s and today, and why did the Scandinavian labor supply decrease less dramatically? Prescott (2004) created a model of the G-7 economies which suggested that the large decrease in the relative labor supply in Continental Europe was due primarily to differences in tax rates. His model allowed him to indirectly calculate the Frisch elasticity of labor supply, the elasticity of hours worked with respect to the wage rate given a constant marginal utility of consumption. He found it to be nearly 3, an estimate at the upper end of the literature on labor supply elasticity. This value implied that individuals would react strongly to changes in tax rates. Therefore, he found it logical that when tax rates in Continental Europe and the United States were comparable during the 1970s, their labor supplies were similar. As tax rates in Continental Europe increased markedly over the next 25 years, its relative labor supply fell dramatically. While Prescott s model fit the empirical data on the labor supplies in the G-7 countries during both and , it did not fit the data for the Scandinavian countries during those periods. Tax rates were higher in Scandinavia in the mid-1990s than in Continental Europe, yet labor supply was also higher. This apparently contradicted Prescott s prediction that people would react strongly to increases in tax rates by reducing the supply of labor. Prescott

8 8 had made numerous assumptions which might have to be relaxed or modified to allow economists to understand more fully the factors which affect labor supply across additional countries. In this thesis, I seek to develop a more effective and realistic model first by building on Prescott s model and then creating my own dynamic general equilibrium model which avoids several of his oversimplifications. My thesis is organized as follows. Chapter 2 begins with a review of the literature on the causes of the relative changes in the labor supply between the United States, on the one hand, and Continental Europe and Scandinavia on the other. It then reviews literature which provides the foundation for my dynamic general equilibrium model. Chapter 3 develops a simple static model based on Prescott (2004). This model employs a less restrictive utility function, and it allows me to derive an original equilibrium relation between hours worked and the effective marginal tax rate and to estimate the Frisch elasticity of labor supply directly. Using this model, I find that the Frisch elasticity of labor supply in the G-7 countries was This suggests that the vast majority of the change in the relative labor supply between the United States and Continental Europe might indeed be explained by higher tax rates in Continental Europe. My estimate of the Frisch elasticity of labor supply corroborates Prescott s indirect calculation. Chapter 4 explains why a number of the underlying assumptions made in the simple static model are gross oversimplifications. For example, that model did not take into account interactions between economic variables across time periods, and it assumed that all non-military government expenditures substitute on a one-to-one basis for private consumption, implying that there was only one consumption good. I then develop a new dynamic general equilibrium model

9 9 and explain why it represents a significant improvement. My presentation provides only an outline of how my calculations were actually performed. It was necessary to encode the model in Matlab because those calculations were far too complex to be performed by hand. That code is presented in Appendix A. Chapter 5 applies my dynamic model to the baseline case in which I assume that (1) the elasticity labor supply is 1.9, (2) consumption-work complementarity exists, and (3) all tax revenue is spent on lump sum transfer payments to individuals regardless of hours worked. The first two assumptions are based on Hall (2009a), and the rationale for them is discussed in chapter 5. The third assumption is derived from Prescott (2004), and the rationale for it is discussed in chapter 3. Chapter 5 explores how an arbitrary increase or decrease in the tax rate would, in the baseline case, affect economic variables including output, hours worked, the real interest rate, the pre-tax wage, consumption, the capital stock, the economy s discounter, and the rental price of capital. It employs a time horizon of 80 years. An increase in the tax rate is shown to lead to a long run decrease in output, hours worked, consumption, and the capital stock while a decrease in the tax rate produces a long run increase in output, hours worked, consumption, and the capital stock. The largest changes occur during the first year after the changes in the tax rate. Thereafter, the variables gradually approach their long term steady states. The real interest rate and pre-tax wage are unchanged in the long run. Appendix B presents the results spreadsheets for changing the tax rate from 20% to 30% or from 20% to 10% in the baseline case Chapter 5 then examines three related issues. It demonstrates that, starting from an initial postulated tax rate of 20%, each 10% increase in the final tax rate from 0% to 90% has an

10 10 increasingly large effect on the magnitude of the changes produced in output, hours worked, consumption, and the capital stock. It shows that removing the assumption of consumption-work complementarity or assuming a less elastic labor supply of 0.5 reduces the magnitude of these changes. Chapter 6 considers alternative scenarios for the amount that individuals value government spending. The assumption that all tax revenue is spent on lump sum transfers to individuals regardless of hours worked is an oversimplification since it is unrealistic to assume that individuals value all government spending as much as direct transfer payments. Therefore, I consider two other scenarios. The first is the extreme case in which all tax revenue is spent on projects which do not affect the individual s marginal utility of consumption in any way. The second scenario is more complex. It is designed to reflect the possibility that a fraction of government spending might provide utility to individuals by increasing their ability to consume while the remainder might be spent on projects which do not affect the individual s marginal utility. I introduce a variable, μ, which represents the fraction of tax revenue spent on programs which correspond to the individuals consumption preferences. In this scenario, the benefit individuals derive from government spending still does not depend on hours worked, but it does depend on the value of μ. The analysis of the results of the changes in tax rates in these scenarios is based on two competing economic effects: the substitution effect and the income effect. The substitution effect is the change in the amount that individuals work given a change in the after-tax wage, holding the marginal utility of consumption constant. The income effect is the change in the

11 11 amount that individuals work because changes in the tax rate make them richer or poorer affecting their marginal utility of consumption. Because in chapter 5 I assumed that all government revenue was spent on lump sum transfers to individuals regardless of hours worked, the income effect was absent. Regardless of the tax rate, tax revenue was fully rebated to individuals. No income was lost, so only the substitution effect influenced hours worked. A lower after-tax wage reduced the incentive to work (marginal utility of consumption being held constant) because individuals would increase their consumption less by working an additional hour. Therefore in chapter 5, increasing the tax rate always led to a decrease in output and hours worked and decreasing the tax rate always led to an increase in output and hours worked. When some or all of government revenue is spent on projects which do not affect the individual s marginal utility of consumption, the income effect becomes relevant. This is because a tax increase will make individuals poorer. When individuals are able to consume less, each unit of consumption is more valuable, so a tax increase will make individuals value each unit of consumption more. Their marginal utility of consumption will increase. Thus, there will be two opposing effects of a tax increase on labor supply when only a fraction of tax revenue is spent on programs which correspond to the individuals consumption preferences. The substitution effect will tend to make individuals work less and the income effects will tend to make them work more. The smaller the value of μ, the greater the income effect. For small values of μ, the income effect dominates in my model, so an increase in the tax rate will cause an increase in labor supply and output. For larger values of μ, the income effect is smaller and the substitution effect dominates, so an increase in the tax rate will cause a decrease in labor supply and output. For a tax decrease, the results are the reverse.

12 12 Chapter 7 explores the effects of tax rate changes on the utility, or welfare, of individuals. First, it presents the changes in welfare produced by changes in the tax rate over the course of an 80 year lifetime. Then it examines the difference between the initial and final steady states. The latter expresses the full impact of the tax changes, but it does not take into account the intervening changes between years 2 and 79, and so it overestimates actual welfare changes which result. This chapter concludes by considering how altering assumptions about the consumption-work complementarity, the elasticity of labor supply, and the nature of government spending would affect welfare changes over the 80 year lifetime. In my baseline case, increasing tax rates reduces welfare and decreasing tax rates increases welfare. This occurs regardless of assumptions made about how individuals value government spending, about whether consumption-complementarity exists, or about whether labor supply is relatively elastic or inelastic, although these assumptions do affect the magnitude of the changes. To provide a more intuitive sense of the magnitude of the welfare changes which I calculate, I consider consumption change equivalents and hours worked change equivalents. Consumption change equivalents express how much more or less consumption an individual would need to have if he continued to work as much as he had at the initial tax rate to attain the utility he actually achieves when the tax rate is increased or decreased. Hours worked change equivalents express how much more or less an individual would need to work if he continued to consume as much as had did at the initial tax rate to attain the utility he actually achieves when the tax rate is increased or decreased. An important finding in this chapter is that the lower the value of μ, the greater the magnitude of the change in welfare caused by a given change in the tax rate. While a low value of μ may imply that an increase in the tax rate will increase output and labor supply, welfare will

13 13 be reduced dramatically when μ is small. In fact, for a given tax rate increase, the higher the value of μ, the smaller the decrease in welfare despite the fact that output and labor supply will be more negatively affected when μ is large. Conversely, for a given tax rate decrease, the higher the value of μ, the smaller the increase in welfare despite the fact that output and labor supply will be more positively affected when μ is large. With this background, chapter 8 returns to the original research questions. I consider eight countries: the United States, the four largest European economies (France, Germany, Italy, and the United Kingdom), the two largest European economies with very high deficit to GDP ratios (Greece and Spain) and the largest Scandinavian economy (Sweden). My first goal is to use my model to predict how much these countries would have to raise taxes in order to balance their budgets within one year. I examine various scenarios incorporating different assumptions about government spending to determine how such assumptions would affect the size of the necessary tax increases and the resulting effects on output and labor supply. I pursue the case of the United States in greater depth, presenting the Laffer Curves which my model generates under different assumptions about government spending. The Laffer curve generated when μ = 0 is particularly interesting because it does not reach a peak before the marginal tax rate is Finally, I compute how much the United States would have to raise taxes in order to balance its budget in the long run, taking into account both the short term and the long term impact of tax rate increases on output. Regardless of whether tax revenue is assumed to be spent on lump sum transfers to individuals regardless of hours worked or whether individuals value government spending ½ as much as their own consumption (μ = 0.50), all countries except Sweden, which has only a small budget deficit, are either incapable of achieving a balanced budget in one year through tax

14 14 increases alone or would suffer dire economic consequences as a result of the necessary tax increases. If a reduction in government spending accounts for 50% of the budget fix and tax revenue is assumed to be spent on lump sum transfers to individuals regardless of hours worked, Germany, Greece, and Spain could not balance their budgets. Of the remaining countries, only Sweden could avoid serious economic repercussions. If the reduction in government spending accounts for 50% of the budget fix and if individuals value government spending ½ as much as their own consumption (μ = 0.50), 7 of the 8 countries can balance their budgets. Greece s budget deficit is so severe and its current tax rates are so high that even under these conditions, it cannot balance its budget. All the countries which can balance their budgets would experience reductions in output, and Germany and Spain would face devastating hardship. In the case of the United States, whose 2009 deficit to GDP ratio was 11% and whose debt to GDP ratio was 53%, I demonstrate that assuming no consumption-work complementarity or a less elastic labor supply would lead to a lower values for tax rate necessary to balance the budget and smaller consequent decreases in output and hours work. The greater the share of the budget fix made up by decreases in government spending, the smaller the necessary increases in the tax rate and the smaller the decreases in output and hours work. The smaller the value of μ, the smaller the necessary increase in the marginal tax rate to balance the budget, and the smaller the decreases in output and hours worked. In my baseline case, I find that for the United States even to achieve a deficit to GDP ratio of 3% without a reduction in spending would require such a large increase in its tax rate that a depression would result. Furthermore, despite the relatively low marginal tax rate of the United States, it would be impossible for it to achieve a budget surplus of 3% simply by raising tax rates without reducing government spending. Because my model demonstrates that only

15 15 about 90% of the decrease in output which occurs in a country over an 80 year lifetime is experienced during the first year after a tax rate increase, I find that the long run fall in output caused by a tax increase would be even larger than the short run fall in output. Therefore, still larger tax increases would be necessary to achieve balanced budgets in the long run, causing greater economic harm. For example, although the United States could balance its budget in one year under the baseline assumptions if it were willing to endure the resulting economic hardship, it would be impossible for it to balance its budget in the long run simply by raising taxes. Chapter 10 presents my conclusions and suggests potential avenues for further research.

16 16 Chapter 2: Literature Review I begin by reviewing literature which analyzes four aspects of the relative changes in the labor supply between the United States, on the one hand, and Continental Europe and Scandinavia on the other. The first is the effect of tax rates on the relative labor supplies. The second is the way that patterns of government expenditures help explain why over the past two decades relative labor supply has been higher in Scandinavia than in Continental Europe despite higher taxes. The third is the influence of labor unions, labor market regulations, and output per hour worked on labor supply. The fourth is the impact of the level of income inequality on labor supply. I then review literature which provides background for Hall (2009a) since this paper provides the foundation for the dynamic model I develop in this thesis Labor Supply Across Countries The Effect of Tax Rates Prescott (2004) spurred intense debate in the economic community over the impact of differences in tax rates on the observed differences in relative labor supply across countries. He defined labor supply as hours worked per person aged in the market sector. He only counted those hours which resulted in taxed labor income. Paid vacations, sick leave, holidays, and time spent working in the underground economy or at home were specifically excluded. Prescott created a model for the G-7 countries following standard macroeconomic theory and analyzed a stand-in household that faced a decision over how much to work and how much to consume. He derived an equilibrium relation between hours worked and the effective

17 17 marginal tax rate on labor income. He explained the way he estimated tax rates and justified his why his estimates were reasonable. One of his key simplifying assumptions was that all tax revenue, except for that used for pure public consumption, was returned to households in lump sum payments independent of household income or hours worked. He defended this assumption by asserting that most public expenditures in the G-7 countries are substitutes for private consumption. This assumption allowed his model to consider only one consumption good. Prescott noted that although this was reasonable for the G-7 countries, it was not valid if a broader set of countries was to be considered. For example, in Scandinavia greater government expenditures on programs like child care for working parents necessitated treating some publicly provided goods separately from privately consumed goods. Rogerson (2006) extended Prescott s analysis to include multiple consumption goods. His paper is discussed later in subsection In Prescott s analysis of the G-7 countries during the periods of and , he found the Frisch elasticity of labor supply was nearly 3, implying that individuals would react strongly to changes in tax rates and that the vast majority of the differences in labor supply between the United States and Continental Europe could be explained by the higher taxes in Continental Europe. He noted that his model estimated the labor supply among the G-7 countries very well during the mid-1990s and quite well during the early 1970s. Prescott s finding of an elasticity of labor supply of nearly 3 is at the extreme upper end of values found in the literature. Many economists believe that the elasticity is actually much lower. Pistaferri (2003) based his estimate of labor supply elasticity on data on how workers expectations of their wages change as tax rates change. He calculated the Frisch elasticity of labor supply to be Studying the decline in hours worked among lottery winners, Kimball and Shapiro (2003) estimated the Frisch elasticity of labor supply to be about

18 18 one. Mulligan (1998) found the elasticity to be not much greater than one. His work is controversial, however, because, unlike most authors, he included data on older workers who earn lower wages, work fewer hours, and tend to be more responsive to changes in tax rates. Rogerson and Wallenius (2008) sought to explain the variation which exists in estimates of the Frisch elasticity of labor supply by dividing the relevant papers into two groups. Those which use microeconomic models typically calculate elasticities ranging from 0.05 to Those papers that use macroeconomic models typically find values between 2.25 and Low values derived from microeconomic models would imply that differences in tax rates between the United States and Continental Europe could only explain a small part of the differences in the changes in relative labor supply because these values imply that the number of hours worked would not be strongly affected by changes in tax rates. The high values derived from macroeconomic models, however, would imply that differences in tax rates between the United States and Continental Europe could explain most of the differences in the changes in relative labor supply. In the papers considered above, home production was not considered to contribute to labor supply. Olovsson (2009) pointed out, however, that while market work per person is about 10% higher in the United States than in Sweden, the total number of hours worked differs by only 1% when hours worked at home are included. He developed a model which demonstrated that differences in tax rates could account for most of the discrepancy in market work and home production between the United States and Sweden. He noted that taxes affect the amount of home production because service taxes raise the price of market-produced services while labor taxes reduce the economic return to market work, increasing the relative return to home production. Because tax rates are much higher in Sweden than in the United States, there is a

19 19 greater incentive for Swedes to substitute home production for market work. Freeman and Schettkat (2005) came to similar conclusions. They found that while Americans work significantly more hours per week in the market sector than Europeans, Americans and Europeans work about the same number of hours per week when home production is included. Their analysis indicated that higher taxes in Europe played a key role in creating this situation The Effect of the Way that Government Spends its Revenue While tax rates definitely affect labor supply, it seems likely that other factors contribute to the differences in relative labor supply among the United States, Continental Europe, and Scandinavia. One of these factors is the pattern of government expenditures, which helps account for the fact that labor supply is higher in Scandinavia than it is in Continental Europe despite higher tax rates in Scandinavia. Rogerson (2006) explored this issue by employing a version of the standard neoclassical growth model which takes into account different types of government spending. Instead of making Prescott s assumption that all government expenditures are lump sum transfers independent of hours worked, Rogerson divided government expenditures into four categories, depending on whether they were used (1) to finance lump sum transfers independent of hours worked, (2) to hire workers at market wages who produced nothing which households valued, (3) to subsidize consumption, or (4) to subsidize leisure. He found that taxes which funded government expenditures to hire unproductive workers or to subsidize consumption had no effect on hours worked, but taxes expended on lump sum transfers and subsidies for leisure had a negative effects on hours worked. Those spent on subsidies for leisure had the largest effect. Rogerson concluded that it is essential to consider how the government spends its revenue in order to determine the effect of tax rates on labor supply because different types of expenditures lead to different effects.

20 20 Rogerson (2007) applied this analysis to the anomaly of Scandinavia. Scandinavia had both higher tax rates and higher labor supplies than Continental Europe in the mid-1990s. This appeared to contradict the theory that higher tax rates lead to lower labor supply. Rogerson found, however, that the elasticity of labor supply varies among countries depending on their patterns of government expenditures. If the elasticity of labor supply in Scandinavia is low enough relative to that in Continental Europe, this could explain how Scandinavian countries could simultaneously have higher tax rates and higher labor supplies. Scandinavian taxes for government expenditures would have a less negative effect on labor supply than taxes in Continental Europe. By assuming more than one consumption good in the economy, Rogerson was able to distinguish between government transfers to individuals regardless of hours worked, government transfers affected by hours worked, and government transfers which did not affect the individual s marginal utility of consumption. If higher taxes fund disability payments which can be received only by an individual who is not working, for example, they might have a strongly negative effect on labor supply. On the other hand, if the higher taxes were used to fund day care of children of working parents, they might have a much smaller adverse effect on labor supply. Rogerson also pointed out that higher taxes would lead to larger differences on the labor supply in activities for which there are strong non-market substitutes. This point was also stressed by Davis and Henrekson (2005). These authors considered, for example, the case of a family who wished to paint their home, an activity for which there is a strong non-market substitute. If the family hires professional painters, the transaction is subject to taxation. If it paints the house itself, taxes are avoided. The higher the taxes, the greater the incentive to avoid

21 21 the market alternative. Other activities, such as auto production, are difficult to carry out outside the market sector. The effect of higher tax rates on labor supply in such activities would be expected to be smaller. Building on this analysis, Rogerson explained how Scandinavian countries could have higher marginal tax rates and still have higher labor supply. Compared to Continental Europe, they have substantially higher rates of government employment. This government employment often serves as an implicit transfer to the government employee, although it does not affect the marginal utility of consumption for the vast majority of households. Higher government employment in the service sector in Scandinavia provides a larger transfer of market services to households, increasing hours worked in the market sector. For example, Rogerson noted that Scandinavian countries spend a much higher percentage of government revenue on services such as child and elderly care. Such jobs often have strong non-market substitutes. The same work which counts toward labor supply in Scandinavia when it is performed by government employees does not count in Continental Europe, where it is often completed outside of the market sector. Rosen (1996) noted that child care subsidies have had an especially large impact on labor supply by encouraging labor force participation by women in Sweden. With Sweden s high marginal tax rates, it would be difficult for these mothers to work in the market sector and purchase child care in the private market. Child care subsidies, however, tend to offset this income tax penalty. Rosen noted that such child care subsidies are a major reason why female participation in the labor market is much higher in Scandinavia than in Continental Europe and thus a major reason why hours worked in Scandinavia are higher than they might otherwise be.

22 The Effect of Labor Unions, Labor Market Regulations, and Productivity Differences in the power of labor unions, the stringency of labor market regulations, and relative productivity also affect relative labor supply across countries. Alesina et al. (2005) argued that differences in the power of labor unions and in the stringency of labor market regulations, rather than differences in tax rates, explain the majority of the differences in labor supply between Continental Europe and the United States. They noted that the strength of unions increased dramatically in Continental Europe during the 1970s and 1980s, the period which witnessed the large decrease in its relative labor supply. They cited Alesina and Glaesar (2004), which argued that American racial fractionalization and European political instability led to Americans being much less open to the socialist/marxist left than Europeans and therefore less supportive of organized labor. Alesina et al. argued that because of their power, unions in Continental Europe were able during economic shocks to push successfully for a reduction in hours worked as an alternative to increased unemployment, employing slogans like work less work all. This might not have been an optimal response to the economic conditions, but union power often rested on the size of union membership, making such policies appealing to them. Hunt and Katz (1998) pointed out that the reduction of the standard workweek did not lead to any significant increase in overtime work in Europe. Therefore, the diminished standard workweek reduced labor supply. Alesina et al. pointed out that unions demanded higher wages to compensate for lower hours worked and keep total income constant. This led to even sharper declines in labor supply because it was difficult for employers to increase hourly wages without making cuts in total hours of employment. Nickell (1998) noted that unions also succeeded in lobbying for much

23 23 higher unemployment compensation in Europe, which reduced the marginal loss of consumption for the unemployed. This made unemployment less undesirable. It reduced the incentive to seek employment and pressured employers to raise wages in order to provide incentives to work. The resulting higher wages reduced the incentive for employers to hire more workers. Rogerson (2006) pointed out another reason for the decrease in the labor supply of Continental Europe and Scandinavia relative to that of the United States. As countries develop, individuals tend to work less. More advanced nations have lower labor supplies because their citizens can achieve a similar standard of living while working fewer hours and enjoying more leisure. As consumption increases, the marginal utility of consumption decreases. Individuals thus have less incentive to work longer hours when they receive higher wages. In the early 1970s, output per hour worked in Continental Europe and Scandinavia was significantly lower than in the United States. By the mid-1990s, productivity in Continental Europe and Scandinavia had caught up (Prescott 2004, Pilarski 2008). Thus, in the early 1970s, Continental Europeans and Scandinavians had an extra incentive to work because of their lower output per worker. By the mid-1990s, this difference no longer existed. If Continental Europeans and Scandinavians worked about as much as Americans when they had an extra incentive, they would be expected to work less without it. According to Pissarides (2007), one of the reasons for the dramatic decrease in the relative labor supply of Continental Europe over this period is that the comparison is between the time of exceptional economic and productivity growth which Europe experienced from the end of World War II to the early 1970s and the normal steady state of the mid-1990s.

24 The Effect of Income Inequality A final factor contributing to the fact that Americans work more than Europeans is that there is greater income inequality in the United States. As Formby et al. (2004) pointed out, the distribution of wages is more even, for example, in Germany than in the United States. In addition, government transfers from the wealthy to the poor are greater in Germany, so disposable income inequality is even less in Germany relative to the United States than wage income inequality (Gottschalk and Smeeding 1997). Bell and Freeman (2001) explained the fact that Americans work more than Germans by analyzing forward-looking labor supply responses to differences in earnings inequality. They argued that workers often choose additional current hours of work because they hope to gain promotions and higher pay in the future. Since earnings are more unequally distributed in the United States than in Germany, extra work has a greater potential payoff in the United States, and this leads to more hours worked. According to Bell and Freeman, the argument that greater earnings inequality leads to more hours worked rests on three premises: (1) that individuals can rise in the percentile distribution of earnings by working more hours, (2) that individuals base how much they work currently on assumptions they make about how much they expect to earn in the future as a result of their current work, and (3) that the greater the inequality of income, the larger the marginal change in an individual s earnings given the same change in his position in the percentile distribution of income. Using data on earnings and hours worked in the market sector in the United States and Germany, they found (1) that the greater the number of hours an individual worked, the greater his expected earnings in both the United States and in Germany, but with a greater expected benefit in the United States and (2) that in both countries full time workers work more hours in occupations which have greater wage inequality. This provided evidence that that

25 25 the difference in income inequality between the United States and Germany was a significant factor in the difference in hours worked in the market sector between the two countries Background of the Dynamic General Equilibrium Model The Neoclassical Starting Point Hall (2009a) sought to estimate the output and consumption multiplier effects which quantify how much output and consumption change given an increase in government purchases. Hall initially developed a static neoclassical model with a standard macroeconomic utility function in which individuals receive positive utility from consumption and negative utility from work. In his model, technology was estimated using the Cobb-Douglass production function, and the real wage was assumed to be determined by the marginal product of labor. Employing this model, Hall calculated the output multiplier to be 0.4 and the consumption multiplier to be These values differ significantly from the empirical data which indicate that the output multiplier is about 1 and the consumption multiplier is about 0. Recognizing this, Hall considered additional factors to make his model more reflective of reality. The first was the endogenous markup of price over cost, which reflects the fact that firms in the real world have market power. Assuming sticky prices, this power is higher in economic slumps and lower in booms. Markups are thus countercyclical, as Rotemberg and Woodford (1992) demonstrated. Augmenting his neoclassical model with a constant-elasticity relationship between markup and output, Hall computed the output multiplier to be 0.5 and the consumption multiplier to be -0.5, a step in the right direction. He then considered unemployment and the employment function, noting that many general equilibrium models have difficulty explaining fluctuations in labor supply if they do not

26 26 explicitly consider unemployment. Monica Merz (1995) and David Andolfatto (1996) made significant progress by building on the work of Mortensen and Pissarides (1994) by explicitly allowing for unemployment in otherwise neoclassical models. Blanchard and Galí (2007) succeeded in introducing unemployment in the New Keynesian Model. Hall (2009b) built on the Mortensen-Pissardes model by allowing for a broad variety of bargaining solutions between jobseekers and employers and replacing their linear preferences with standard macroeconomic preferences. Hall (2009a) then used an employment function, which he had previously developed in Hall (2009b). This function incorporates two components. The first is the fraction of the population which is employed as a result of the interaction between jobseekers and employers. The second is the Frisch supply function for hours worked per employed worker, which explicitly takes unemployment into account. Chetty et al. (2011) illustrated why, in measuring labor supply elasticities, it is important to consider both how much people respond to changes in the tax rates conditional on employment and how many people choose to be employed when tax rates change. Using his employment function, Hall calculated a labor supply elasticity much higher than that which he had found using the standard neoclassical model. This value was more in line with labor supply elasticity calculated by Kydland and Prescott (1982). Although that value had been criticized for by Mankiw et al. (1985) and Eichenbaum et al. (1988) as being too high, Hall stated that in light of recent work, it appeared reasonable. With this basis, he reestimated the output multiplier to be 0.8 and the consumption multiplier to be -0.2, values even closer to the empirical data. The last factor which Hall added to his static model was the consumption-work complementarity, which reflects the fact that as an individual works more hours in the market

27 27 sector, his marginal utility of consumption rises. As an individual has the less time for home production, he places greater value on consumption in the market. Aguiar and Hurst (2005) and Hurst (2008) analyzed consumption patterns upon retirement. Their data supported the notion of complementarity between consumption and work. There is a significant drop in consumption of goods and services in the market when individuals cease\ to work, apparently because they have the opportunity to spend more time on home production and therefore have less need to make purchases in the market. Browning and Crossley (2001) and Low et al. (2008) studied declines in consumption in the market during periods of unemployment. Their work also supported the notion of consumption-work complementarity. Employing the preferences used in Hall and Milgrom (2008) to account for consumption-work complementarity, Hall calculated an output multiplier of 0.97 and a consumption multiplier of These values were consistent with empirical evidence Dynamic Modeling: Building on his static model, Hall then constructed a dynamic model assuming that individuals smooth consumption over their lifetimes instead of simply consuming their present incomes. This dynamic model resulted in lower output multipliers for government purchases because temporary government purchases were offset by future tax increases, and individuals prepared for such increases by cutting current consumption to smooth their consumption by saving to meet the anticipated tax increases. In his model, Hall employed standard treatment of capital adjustment costs, the rental price of capital, the equivalence of capital demand and capital supply, the law of motion for capital, the economy s discounter, and the Euler equation. He also assumed that after an initial increase in government purchases, government spending would be reduced each year thereafter until it asymptotically approached the pre-shock level and that the

28 28 capital stock after government spending would also return to the pre-shock level. Using this model, Hall computed an output multiplier of 0.98 and a consumption multiplier of He found that if he removed his assumptions about endogenous markup or assumed a lower labor supply elasticity, this had a large effect on his calculations, but if he assumed no consumptionwork complementarity, this had only a small effect. Many economists have taken issue with the idea that all individuals smooth consumption over their lifetime. One of Keynes s major contributions to macroeconomics was the idea that current consumption depends heavily on current income. Nevertheless, it appears reasonable to assume that some individuals have full access to capital markets and can smooth lifetime consumption although other may be credit constrained and thus consume current income. Galí et al. (2007) followed this logic in employing a standard New Keynesian model for government purchases. They considered a population in which a fraction of individuals, λ, consumed all of its labor income while the remainder smoothed lifetime consumption. The output and consumption multipliers depend strongly on the value of λ. Higher values lead to much larger multipliers. López-Salido and Rabanal (2006) built on the leading New Keynesian model of Christiano et al. (2005), employing a similar assumption about the existence of two groups of individuals, one of which consumes current income while the other smoothes lifetime consumption. López-Salido and Rabanal agreed that the output and consumption multipliers depend strongly on the value of λ, finding that a higher λ leads to much larger multipliers. When Coenen and Straub (2005) employed the model of Smets and Wouters (2003) and assumed two groups, they confirmed that the value of λ affected their calculation of the multipliers, although their multipliers were much lower than those computed by Galí et al. and López-Salido and Rabanal.

29 29 Chapter 3: A Static Model of Tax Rates and Labor Supply Before developing a dynamic general equilibrium model to analyze how tax rates and the patterns of government spending affect labor supply, it is useful to employ a simpler model and to understand the economic theory behind it. In this chapter, I develop such a model based on the work of Prescott (2004). I present background information for this model, discuss the model s theoretical framework, and derive its key equilibrium relation. Finally, I explain the results obtained using this model Background Prescott (2004) created a static model of the G-7 economies which explained that the change in the relative labor supply between the United States and Continental Europe between the early 1970s and the mid-1990s was largely a result of differences in tax rates. His model employed a utility function which established an equilibrium relation between hours worked and the effective marginal tax rate. It allowed for an indirect calculation of the Frisch elasticity of labor supply. Prescott found this elasticity to be nearly 3, an estimate at the upper end of the range of those in the macroeconomic literature. My simple static model builds on Prescott s work, but it employs a less restrictive utility function. This allows me to derive an original equilibrium relation between the effective marginal tax rate and labor supply and to directly calculate the Frisch elasticity of labor supply to be This model would tend to support Prescott s conclusion that the vast majority of the differences in labor supply between the United States and Continental Europe can be explained by differences in the marginal tax rate.