Estimating high-income tax elasticities using sub-national variation in tax rates Kevin Milligan Vancouver School of Economics University of British Columbia Michael Smart Department of Economics University of Toronto January, 2014 Results are preliminary. Comments welcome. Milligan and Smart: High Income Taxation 1
How do individuals respond to higher income tax rates? Real responses Labour supply consumption/savings. Evidence variable, but effects in many cases limited. Shifting responses. Timing Accounting / form / arbitrage Some effects very large. Milligan and Smart: High Income Taxation 2
Who Responds? \ Real: Anyone can adjust labour supply or consumption / savings. Do hyper-competitive top lawyers and hedge fund workers really cut hours worked in response to taxes? Shifting / avoidance Shifting responses assisted by expensive advice, costly financial structures. Advantage goes to higher incomes: Do we expect those with access to the very best tax advice to fail to heed it? Milligan and Smart: High Income Taxation 3
Veall (2012) CEA Presidential Address Milligan and Smart: High Income Taxation 4
Any role for taxes in the ‘surge’ ? In United States: Time series lines up fairly well with Reagan tax cuts in early 80s and especially TRA 1986. Piketty and Saez are skeptical though…. In Canada: Veall: taxes don’t seem to matter a lot in explaining the trends Milligan and Smart: High Income Taxation 5
Questions posed in this paper Q: What impact have taxes had on high income trends? Q: How much success would taxes have in pushing against high income trends? Milligan and Smart: High Income Taxation 6
Preview of findings We find substantial elasticities, concentrated at higher incomes o Elasticities for P99.9 over 1.0; 4 times that of P99-P99.9. Veall is right —taxes don’t explain the ‘surge very well. o But, the tax cuts of the 2000s can account for about one third of top income share growth in that decade. Some evidence that Alberta might be different; capital and earned income affected. o Starts to uncover some hints about mechanisms. Milligan and Smart: High Income Taxation 7
Outline of talk: 1. Brief review of other evidence 2. Quick tax backgrounder 3. Empirical strategy / Identification 4. Data 5. Main estimation results 6. Policy implications 7. Extensions: Alberta and composition of income. Milligan and Smart: High Income Taxation 8
Other evidence: Survey by Saez Slemrod and Giertz 2012: While there are no truly convincing estimates of the long-run elasticity, the best available estimates range from 0.12 to 0.40. Proceeding mechanically, at the approximate midpoint of this rate — an ETI of 0.25 Canadian evidence: Sillamaa and Veall (2001) 1.67 at the top Gagne, Nadeau, and Vaillancourt (2000) over 3.0! Saez and Veall (2005) Big / small Department of Finance (2010) around 0.6 Milligan and Smart: High Income Taxation 9
Canadian Personal Income Taxes Two important features: 1. Tax base is co-occupied by federal government and provinces. In 2009, 39.6% of the $189B in income tax went to provinces. 2. Tax Collection Agreements impose common tax base on provinces. Queb ec not ‘agreeing province’; differences are fairly small. Milligan and Smart: High Income Taxation 10
Tax brackets in BC 2010: Combined federal / provincial brackets and rates Taxable Income BC Tax Federal Tax Combined Tax Up to $35,859 5.06% 15% 20.06% …up to $40,970 7.7% 15% 22.7% …up to $71,719 7.7% 22% 29.7% …up to $81,941 10.5% 22% 32.5% …up to $82,342 10.5% 26% 36.5% …up to $99,987 12.29% 26% 38.29% …up to $127,021 14.7% 26% 40.7% …over $127,021 14.7% 29% 43.7% Canada features almost no progressivity for incomes in the ‘top 1%’ over our time period. Relieves us a bit from worrying about how to measure tax rate for those in top 1% Milligan and Smart: High Income Taxation 11
Figure 1: Top Tax Rates for 2013 by Province Source: Canadian Tax and Credit Simulator Milligan and Smart: High Income Taxation 12
Figure 2: time path of top rate Source: Canadian Tax and Credit Simulator Milligan and Smart: High Income Taxation 13
Empirical Strategy Interested in the following economic relationship: ( ) With some manipulation, this can be written in the form of a group share of total income , in log form to directly estimate elasticity e : ( ) Why use shares? Saez (2004): Against: not the same people in the share each year For: estimating shape of income distribution for tax policy analysis, is ok. Against: time series identification is very tricky. Milligan and Smart: High Income Taxation 14
Saez, Slemrod, and Giertz 2012 JEL: researchers should be seeking better sources of identification; for example, parallel income tax systems that differentially affect taxpayers over a long period of time. Milligan and Smart (2014): OK! We can compare across provinces: ( ) Milligan and Smart: High Income Taxation 15
What do we get out of introducing multiple jurisdictions? 1. Same tax base The elasticity e is a function of a given tax system. Comparing across tax systems and you ’ re not measuring the same thing. 2. Control for common time effects in a flexible way. We can have our time effects and even try provincial polynomial trends. … but we ’ re not the first on this: Atkinson and Leigh (2010) compare across 5 anglo-sphere countries Finance (2010) do what we do with Canada. Milligan and Smart: High Income Taxation 16
Two more empirical manipulations 1. We use the log share ratio as extra control controls for trends in complement group to g : Using income shares ( ) ⁄ ) ( ( ) 2. We instrument for using a simulated tax rate based on an income level common across all provinces and years This is standard in the literature, but almost irrelevant here since top tax bracket rarely above P99 cutoff. Milligan and Smart: High Income Taxation 17
Data Main source of data is the CANSIM high income database (204-0002) Ultimate source is Longitudinal Administrative Database (taxfiler data) 20 percent sample of Canadian taxfilers between 1982 and 2011. We use ‘total income ex cluding capital gains’ definitio n Taxes are calculated using Canadian Tax and Credit Simulator (CTaCS) Available 1962 to 2012 Don’t have to worry about refundable tax rates or many other issues for these high-income people. Simulated all provinces/years using common 2000 income distribution Also use 1991- 2010 ‘Income Statistics’ put out by CRA Taxes are calculated using Canadian Tax and Credit Simulator (CTaCS) Reports total amount and total people for series of nominal categories. Milligan and Smart: High Income Taxation 18
Table 1: Top Income Thresholds 1988 2011 P95 P99 P99.9 P95 P99 P99.9 Canada 88,520 154,133 444,893 105,229 203,656 677,430 NL 71,667 108,810 98,039 168,094 PE 70,031 114,046 81,715 139,722 NS 78,212 129,099 333,792 90,266 154,588 399,831 NB 74,121 116,173 86,282 147,010 363,395 QC 81,648 133,190 316,612 91,140 169,649 505,254 ON 95,720 179,495 579,882 108,047 215,316 778,189 MB 78,376 125,336 310,394 91,917 161,098 469,595 SK 79,357 125,663 288,796 107,269 180,240 507,878 AB 92,284 160,515 416,422 138,945 281,096 1,044,419 BC 89,502 157,897 457,001 102,411 190,151 650,418 Source: CANSIM 204-0002 Milligan and Smart: High Income Taxation 19
Figure 3: Top Income Shares 1988-2011 Source: CANSIM 204-0002 Milligan and Smart: High Income Taxation 20
Figure 4: Top Income Share Index 1988-2011 Source: CANSIM 204-0002 Milligan and Smart: High Income Taxation 21
Results overview Results focused on the log-income share equation; main parameter of interest is the elasticity of reported income. Start with a basic ‘ time series ’ specification like the literature to show value of having provinces there. Try out: o Different fractiles o Different time periods o Different income definitions Milligan and Smart: High Income Taxation 22
Figure 5: Basic Scatter Plot Milligan and Smart: High Income Taxation 23
Table 2: Comparing Time Series to Prov-Year (1) (2) (3) (4) (5) National Time Series Province-Year Data Add No Add US Province Add US time Controls top 1% controls top 1% controls Observations 24 24 240 240 240 R-squared 0.473 0.849 0.71 0.875 0.941 Log Actual MTR 2.032*** 0.485 1.664*** 0.505* 1.082* [0.319] [0.322] [0.361] [0.241] [0.496] Log US Top One - - Percent Share 5.690*** 5.130*** [0.809] [0.492] Source: CANSIM Series 204-0002 and CTaCS Milligan and Smart: High Income Taxation 24
Table 3: Different specifications (1) (2) (3) (4) (5) (6) Add Basic IV income Unweighted Log share Linear Quadratic Dependent Provincial Provincial Variable Trends Trends Observations 240 240 240 240 240 240 R-squared 0.941 0.97 0.949 0.97 0.975 0.988 Log Actual MTR 1.068** 0.689*** 0.723** 0.640*** 0.794** 0.510* [0.441] [0.238] [0.293] [0.210] [0.323] [0.264] Log Total 0.729*** 0.805*** 0.631*** 0.903*** 1.523*** Income [0.0767] [0.0341] [0.0690] [0.179] [0.209] Source: CANSIM Series 204-0002 and CTaCS Milligan and Smart: High Income Taxation 25
Recommend
More recommend
