Econ 551 Government Finance: Revenues Fall 2018

Transcription

1 Econ 551 Government Finance: Revenues Fall 2018 Given by Kevin Milligan Vancouver School of Economics University of British Columbia Lecture 6: Tax Incidence ECON 551: Lecture 6 1 of 42

2 Agenda: 1. Overview 2. Basic Analysis 3. Harberger (1962) 4. Fiscal Incidence 5. Applied Policy Analysis Practice ECON 551: Lecture 6 2 of 42

3 Overview Economics is at its best when it offers important insights that contradict initial, casual impressions. The theory of tax incidence provides a rich assortment of such insights. Larry Kotlikoff Larry Summers Handbook of Public Economics, Volume 2, Chapter 16, ECON 551: Lecture 6 3 of 42

4 Overview When taxes are levied, who really pays them? The producer? The consumer? Someone else? Tax incidence lies behind several policy debates in Canada in recent memory The HST. During the 2011 HST referendum in BC, there was great debate about whether and to what extent the old PST embedded in prices was passed on to consumers. Corporate taxes. Populists seem to believe, at the same time, that taxes on corporations should go up, but also that companies often pass on tax increases to consumers. Dots should be connected Carbon taxes / cap and trade. Very little of carbon pricing affects retail market; most is intermediate production. Therefore, incidence assumptions are critical. o [Q: how do we justify carbon taxes if the Diamond-Mirrlees production efficiency lemma which said not to tax intermediate inputs?] ECON 551: Lecture 6 4 of 42

5 Agenda: 1. Overview 2. Basic Analysis 3. Harberger (1962) 4. Fiscal Incidence 5. Applied Policy Analysis Practice ECON 551: Lecture 6 5 of 42

6 Basic Analysis What is the definition of tax incidence? Fullerton and Metcalfe: analysis of the impact of taxes on the distribution of welfare within a society. Economists draw a distinction between who is legally obligated to remit the tax and those who actually lose economic welfare as a result of the tax. Here are the terms we use: Statutory incidence Economic incidence The legally obligated payer. The person who s welfare changes. Standard result: economic incidence independent of statutory incidence. ECON 551: Lecture 6 6 of 42

7 Passing the incidence of a tax If there is a difference between statutory and economic incidence, the difference is said to be passed on to other economic agents. If a tax on a commodity results in the firm paying lower input prices (e.g. lower profits on capital, lower wages on labour) then we say the tax was passed backwards. If, on the other hand, the tax results in higher prices for the output, we say the tax was passed forwards to consumers. ECON 551: Lecture 6 7 of 42

8 Finding some fuzziness in tax incidence concept Another key question is to dig into what exactly the conceptual experiment is that we have in mind. There are three possibilities: a) Absolute incidence. What is the effect of imposing the tax, assuming that the government just keeps the revenue (or buries it or doesn t use it in any way). b) Balanced-budget incidence. Assume that the government spends the new tax revenue; adjust incidence analysis for the effect of the spending. c) Differential incidence. Reduce some other tax as the tax under consideration is raised. Of course, this then muddies the water so you are analyzing two tax changes at once. So, the question of the incidence of one particular tax is not well-posed. What this makes clear is that the definition of the exercise is not terribly clear. ECON 551: Lecture 6 8 of 42

9 The basic partial equilibrium diagram A tax levied on consumers. Price S 0 tax P 0 P 1 D 0 D 1 Q 1 Q 0 Quantity ECON 551: Lecture 6 9 of 42

10 A tax levied on producers. Price S 1 S 0 P 1 tax P 0 D 0 Q 1 Q 0 Quantity ECON 551: Lecture 6 10 of 42

11 The basic partial equilibrium analysis Notice that the price increase is less than the amount of tax. So, the suppliers eat some of the tax increase and pass some of it on to consumers. Note that a supply shift responding to the tax on supply does the same thing. So, statutory incidence does not matter. Note the dependence of the degree of shifting on the elasticities of demand and supply. If supply is perfectly elastic, then price doesn t go up; suppliers eat the difference. If demand is perfectly inelastic, they pay the full amount. o We can show this more formally ECON 551: Lecture 6 11 of 42

12 The basic partial equilibrium analysis Starting position: supply equals demand at price p: D(p) = S(p) Now introduce tax t, which shifts equilibrium price to p. Assume consumer pays tax. D(p + t) = S(p ) Take derivative of equilibrium with respect to tax rate: D p ( p t + 1) = S p p t. Collect terms: p t = D p ( D p S p ) ECON 551: Lecture 6 12 of 42

13 The basic partial equilibrium analysis Let s do an elasticity trick here: ε D D p p D Substitute these elasticities into previous term: Recognizing D=S gives: p t = p ε S S p p S t = ε D D p (ε D D p ε S S p ) ε D (ε D ε S ) ECON 551: Lecture 6 13 of 42

14 Extensions This analysis was obviously very simple. How could it be extended in interesting ways? General equilibrium analysis This is a method pioneered by John Whalley and others. Typically, you take a very simplified economy (e.g. two sectors, two factors, two outputs). Equilibrium conditions are set (equal returns after tax, risk adjusted in each sector for both labour and capital). The advantage is you can look at general equilibrium effects as the prices and quantities are determined endogenously. The disadvantage is that you have to choose a particular functional form for preferences and technology and other factors. For example, if you assume that capital is mobile internationally (and therefore supplied elastically) then you get very different results than if you assume that there is a fixed amount of capital. ECON 551: Lecture 6 14 of 42

15 Extensions Imperfect competition In models of imperfect competition, you can actually have overshifting. That is, firms increase prices by more than the amount of tax. This happens because firms have some market power and want to try to recoup some of their lost revenue. To do so, they cut back on production and price can rise a lot, depending on the elasticities involved. A lot of empirical work finds evidence of over-shifting. However, in my quick look at it, I m left wondering whether the finding in many of these cases is driven by a bad empirical design or by an actual over-shifting effect. ECON 551: Lecture 6 15 of 42

16 Extensions Lifetime incidence analysis If you evaluate welfare over an entire lifetime rather than just at a point in time, you can get a very different picture. For example, think of a lifecycle savings model. At some stages, you have much income and less consumption. At other stages you have little income but a lot of consumption. Because of this, the distributional impact of a tax could be evaluated differently depending on which age you currently are. In a lifetime analysis, this wouldn t be true. Intergenerational analysis Intergenerational analysis stacks lifetime analyses for different cohorts. This approach was pioneered by Auerbach and Kotlikoff, but Larry Kotlikoff has really pushed it for the last 20 years. This kind of approach is especially useful for tax and benefit policies that have strong intergenerational components such as Social Security or plans for public health spending. ECON 551: Lecture 6 16 of 42

17 Agenda: 1. Overview 2. Basic Analysis 3. Harberger (1962) 4. Fiscal Incidence 5. Applied Policy Analysis Practice ECON 551: Lecture 6 17 of 42

18 Harberger on the incidence of corporate taxes In Harberger (1962 JPE), he sets out a very simple general equilibrium model with two sectors (corporate, non-corporate) two inputs (labour, capital) two outputs (X,Y) He analyzes the impact of corporate taxation on his system. The corporate tax is assumed to be a tax on the return to capital invested in the corporate sector. ECON 551: Lecture 6 18 of 42

19 Harberger on the incidence of corporate taxes What could shift when a tax is imposed? Well, if K-L substitutability is large, then firms might cut back on capital and hire more labour. On the other hand, capital might just shift out of the taxed industry into the untaxed one. The key to the model is that the tax leads to capital leaving the corporate sector and heading to the non-corporate sector in order to equalize after tax returns. Thus, the loss is shared by capital owners in both sectors. That is, capital bears most of the burden of the tax. Capital in both sectors! The elasticities of substitution between labour and capital are just not plausibly large enough to lead to a shift toward more labour in production, which would have shifted the burden onto labour as you rode down the MP L curve. The work of Harberger generated a generation of papers using general equilibrium models. Shoven and Whalley in the 1970s added a c for computable to this to form the well-known modeling technique of computable general equilibrium models (CGE). ECON 551: Lecture 6 19 of 42

20 Agenda: 1. Overview 2. Basic Analysis 3. Harberger (1962) 4. Fiscal Incidence 5. Applied Policy Analysis Practice ECON 551: Lecture 6 20 of 42

21 Fiscal Incidence Analysis: Typically built with: Some assumptions on incidence of different taxes Microsimulation models applying taxes to individuals / firms. Goal is to check out progressivity of tax system: do tax burdens rise? Measure could be income, lifetime income, or consumption. Look only at tax side? Look also at spending? How value national defence? A park? Universities? ECON 551: Lecture 6 21 of 42

22 The Prototype: Pechman and Ockner (1974) This paper was (and still is) the model for a certain type of incidence study. Assumed something on incidence for each tax, then worked through sample of 87,000 people from 1966 tax return data. For example, that the corporate tax was 50% shifted backward to factors of production and 50% forward to prices and consumers. Why 50%? cause they thought it sounded right. [See Fullerton and Metcalfe p.1818 for discussion of P&O.] Of course, they did sensitivity analysis, but the point is they pulled the number out of the air. They included the SS payroll tax but ignored the fact that these taxes pay for benefits down the road. Can be tricky if intergenerational transfers. Also value compared to equivalent pension/insurance scheme. The key result from P&O: tax system is more or less proportional. o Result not sensitive to extreme assumptions for the incidence of different taxes. ECON 551: Lecture 6 22 of 42

23 Canadian examples of fiscal incidence analysis A similar analysis to Pechman-Ockner was undertaken by Gillespie (1980) for Canada. Finds a similar lack of progressivity and even regressivity at the bottom. However, Gillespie uses a net fiscal approach where he looks at the value of benefits less taxes. Other papers have been by Vermaeten, Gillespie, and Vermaeten (1994 CTJ), Ruggeri, Van Wart, and Howard (1994 CTJ). Whalley (1984) looks at Canadian data and shows how sensitive the results of this kind of analysis can be to the assumptions. ECON 551: Lecture 6 23 of 42

24 John Whalley 1984 CJE (Taken from Kesselman and Cheung 2004) ECON 551: Lecture 6 24 of 42

25 John Whalley 1984 CJE As noted by Kesselman and Cheung 2004: 1. for the central case, not a super lot of progressivity. It s not monotonic, and stays pretty much at 35% for the middle 50%. 2. Using very progressive assumptions (e.g. that corporate taxes borne more by capital; sales taxes by producers etc.) you get a super progressive incidence. 3. Using more regressive assumptions, you get super regressive. ECON 551: Lecture 6 25 of 42

26 Dagmar Dyck 2005 Canadian Tax Journal In the style of Pechman and Ockner, but looks both at tax and spending incidence, at different levels of government. Uses a concept referred to as RSA: Relative Share Adjustment index. This tells varies from 0 to 2 and tells us the degree of redistribution relative to a neutral regime (=1.0). So, 1.15 means that a group gets 15% more than they would under a neutral regime. First, let s look at the overall redistribution findings: Fiscal system overall is fairly redistributive. This is in contrast to the Gillespie and Gillespie/Vermaeten studies. Difference is in assumptions. Purchases is pretty flat So are transfers!! Taxes are a big contributor to progressivity Across government levels, it is the federal government that does most of the heavy lifting mostly because they are the biggest players in the income tax. ECON 551: Lecture 6 26 of 42

27 ECON 551: Lecture 6 27 of 42

28 ECON 551: Lecture 6 28 of 42

29 Comments on Dyck 2005: PIT is heavily redistributive. CIT is actually redistributive. (assumptions??) Indirect taxes are the most regressive. This is sales taxes, excise taxes, GST. Payroll taxes are more or less proportional. Overall comments on Dyck style analysis: Assumptions o On expenditure, education allocated just on age, health on sex-age. Not on incomedifferentiated takeup. o Corporate tax incidence solely on dividends and capital gain income; not wages. Cross sectional vs lifetime Statutory vs economic ECON 551: Lecture 6 29 of 42

30 Bird and Smart (2016 CTJ): Look at incidence of consumption taxes, and specifically GST/HST in Canada. Argue that consumption is the right base to assess consumption taxes. Show that you get very different results on the progressivity when using income vs consumption as the base. ECON 551: Lecture 6 30 of 42

31 ECON 551: Lecture 6 31 of 42

32 Summary of Fiscal Incidence: Whalley concludes his 1984 paper by saying that he doesn t mean to be so negative the question of incidence is so important that we ll just have to learn to live with poor data and very sensitive answers. He thought it would be a key thing worked on for years to come. Didn t happen. Instead, as we will see in the empirical papers we study, the trend in more recent work has been to focus on the smaller questions. The mindset is, let s pick something small and manageable where there data are good and we have a hope of answering very well the question, even if the question is very narrow. This is a mindset that pervades a lot of modern empirical work on tax incidence. ECON 551: Lecture 6 32 of 42

33 Agenda: 1. Overview 2. Basic Analysis 3. Harberger (1962) 4. Fiscal Incidence 5. Applied Policy Analysis Practice ECON 551: Lecture 6 33 of 42

34 Some common methods in applied policy analysis We re going to review three methods that are often seen in applied policy work think tanks and governments. These methods require very strong assumptions to make the numbers interpretable. Those assumptions are not always made very clear. Our goal here is to examine the assumptions so you can be better consumers (producers?) of these. 1. Distributional tables 2. Typical taxpayer 3. Tax Expenditures ECON 551: Lecture 6 34 of 42

35 Distributional Table Do a cross tab of the item of interest by income group. Here is capital gains income, from 2002 Census Family SLID PUMF: Income by Has any Avg dollars Cap gains $10K Cap Gain? of Cap Gain share of inc ECON 551: Lecture 6 35 of 42

36 Distributional Table Criticisms 1) Implicitly assumes that statutory burden is economic burden. Corporate taxes never get allocated at all. They do this to avoid making assumptions. But, assuming zero can actually be making a very strong assumption. 2) Annual income, not lifetime. (Example: capital gains bumps you up in income the year of realization.) 3) Ignores excess burden of the tax. 4) Ignores any capitalization of tax into asset prices. ECON 551: Lecture 6 36 of 42

37 Tables of numbers don t make effective political communication So, political communications experts came up with something even more narrow than the distributional table the typical family vignette. Pick out a few families that illustrate what is going on. (How selected?) Easier for non-experts to understand. Obscures details that matter for a more complete understanding by experts. Who is the budget for? o My view: fine with the communications strategy, but be sure also to provide the technical details. Example from 2016 federal budget: ECON 551: Lecture 6 37 of 42

38 Increased Benefits for a Family of Four With $90,000 Example Aveen and Sarita have two children aged 8 and 5. Aveen earned $30,000 and Sarita earned $60,000 in Together, they would have received $3,145 under the current system (for the July 2016 to June 2017 benefit year). In comparison, the Canada Child Benefit program will deliver $5,650 in tax-free payments a net after-tax increase of $2,505. Increased Benefits for a Single Parent With One Child and $30,000 Example Samantha is a single parent with one child aged 3. She earned $30,000 in Samantha would have received $4,852 under the current system (for the July 2016 to June 2017 benefit year). Instead, she will receive $6,400 in tax-free Canada Child Benefit payments a net after-tax increase of $1,548. Increased Benefits for a Family of Four With $120,000 Example Ann and Derek have two children aged 7 and 4. Derek earned $84,000 and Ann earned $36,000 in Together, they would have received $1,901 under the current system (for the July 2016 to June 2017 benefit year). They will receive $3,940 in taxfree Canada Child Benefit payments a net after-tax increase of $2,039. ECON 551: Lecture 6 38 of 42

39 Tax Expenditures Find the annual publication here: Imagine that the government wants to give $100 a year to people with brown hair. They have two ways to do this: a) Cut a cheque from the expenditure side of the budget for this brown hair program. b) Give a special $100 tax reduction from the revenue side of the budget. Designed properly, they do the exact same thing. Is it a tax cut? Is it spending? In recognition of this reality, in the 1970s the concept of the tax expenditure arose. The idea is to take a given tax measure and estimate how much revenue was lost to the tax authority because of the measure. ECON 551: Lecture 6 39 of 42

40 ECON 551: Lecture 6 40 of 42

41 Comments on Tax Expenditures 1) These are normally done assuming no change in behaviour. That is, people do not react to the after-tax change in prices. 2) The foregone tax revenue is calculated by comparing to some benchmark tax system; often left not well defined. What about the Basic Personal Amount? Would we every really have a tax system that didn t feature one? Why use such an odd thing as the ideal benchmark? 3) Each line is estimated separately; assuming all else stays the same. Therefore, you cannot add them. But people do using the tax expenditures report as a menu. 4) Flow or NPV of tax benefits? For RRSPs, it may soon be the case that the flow measure becomes negative as the tax on withdrawals may start to exceed the tax loss on accruing income and contributions. The alternative is a NPV method, but that requires a boatload of sensitive assumptions. 5) Best use is to gain a sense of magnitude for different tax measures, rather than precise menu. Critique: See Brooks (CTJ, 2016). ECON 551: Lecture 6 41 of 42

42 Next Class: More on tax incidence. ECON 551: Lecture 6 42 of 42