INCOME INEQUALITY, SMALL BUSINESS TAXATION AND LOBBYING

Transcription

1 INCOME INEQUALITY, SMALL BUSINESS TAXATION AND LOBBYING Joe Lesica Department of Economics Wilfrid Laurier University November 9, / 28

2 1 - INCOME INEQUALITY ON THE RISE GINI COEFFICIENT OF INCOME INEQUALITY Canada USA GINI COEFFICIENT SOURCE: OECD INCOME DISTRIBUTION DATABASE 2 / 28

3 2A - RECENT TREND: RISING TOP 1% INCOME SHARE TOP 1% (FISCAL) INCOME SHARE Canada USA CANSIM 20 % SOURCE: WORLD WEALTH AND INCOME DATABASE 3 / 28

4 2B - RECENT TREND: RISING TOP 0.1% INCOME SHARE TOP 0.1% INCOME SHARE 12.5 Canada USA 10.0 % SOURCE: CANSIM ; SAEZ (2016) 4 / 28

5 3 - ACCOUNTING FOR (SMALL) BUSINESS INCOME Wolfson, Veall, etc. (2016) - FIGURE 10 - Trends in Top Income Shares 5 / 28

6 MOTIVATION: SMALL BUSINESS TAX POLICY ROLE? 1 Wolfson, Veall, et al. (2016): Retaining income within small businesses plays an important role in income inequality measurement and rise. 6 / 28

7 MOTIVATION: SMALL BUSINESS TAX POLICY ROLE? 1 Wolfson, Veall, et al. (2016): Retaining income within small businesses plays an important role in income inequality measurement and rise. 2 PM Trudeau s tax shelter claim in 2015 election: We have to know that a large percentage of small businesses are actually just ways for wealthier Canadians to save on their taxes... 6 / 28

8 MOTIVATION: SMALL BUSINESS TAX POLICY ROLE? 1 Wolfson, Veall, et al. (2016): Retaining income within small businesses plays an important role in income inequality measurement and rise. 2 PM Trudeau s tax shelter claim in 2015 election: We have to know that a large percentage of small businesses are actually just ways for wealthier Canadians to save on their taxes... To clarify... in Canada, small business takes the legal form of a Canadian Controlled Private Corporation (CCPC) Reduced tax on the first $500, supposed to encourage investments. 6 / 28

9 SMALL BUSINESS RATE (τ) < TOP INDIVIDUAL RATE (t) Recognition that high net worth individuals take advantage of SBT < Top P I T for personal tax avoidance. 30 AB BC MB NB NL % NS ON PE QC SK / 28

10 CONTRIBUTIONS AND RESULTS OF THIS PAPER 1 Provide a theoretical framework for income shifting and SBT increasing top income share. Milligan & Smart [2015] 8 / 28

11 CONTRIBUTIONS AND RESULTS OF THIS PAPER 1 Provide a theoretical framework for income shifting and SBT increasing top income share. Milligan & Smart [2015] 2 Policy of SBT is NOT income inequality neutral. Unambiguously increases the share of income going to the top p% of business owners. (Gordon & MacKie-Mason[1995], Gordon & Slemrod [1998], Sivadasan & Slemrod [2008], Alstadseter et al. [2016]) 8 / 28

12 CONTRIBUTIONS AND RESULTS OF THIS PAPER 1 Provide a theoretical framework for income shifting and SBT increasing top income share. Milligan & Smart [2015] 2 Policy of SBT is NOT income inequality neutral. Unambiguously increases the share of income going to the top p% of business owners. (Gordon & MacKie-Mason[1995], Gordon & Slemrod [1998], Sivadasan & Slemrod [2008], Alstadseter et al. [2016]) 3 Have small businesses organized in a Special Interest Group (SIG) Able to buy a SBT lower than the revenue maximizing one by lobbying. Theoretical mechanism for Chirinko & Wilson [2010] findings. 8 / 28

13 CONTRIBUTIONS AND RESULTS OF THIS PAPER 1 Provide a theoretical framework for income shifting and SBT increasing top income share. Milligan & Smart [2015] 2 Policy of SBT is NOT income inequality neutral. Unambiguously increases the share of income going to the top p% of business owners. (Gordon & MacKie-Mason[1995], Gordon & Slemrod [1998], Sivadasan & Slemrod [2008], Alstadseter et al. [2016]) 3 Have small businesses organized in a Special Interest Group (SIG) Able to buy a SBT lower than the revenue maximizing one by lobbying. Theoretical mechanism for Chirinko & Wilson [2010] findings. 4 Empirically verify the negative relationship between SBT and top income shares in Canada. 5 Empirical evidence that lobbying has significant negative influence on small business tax rates across Canadian provinces. 8 / 28

14 MODEL - OPTIMAL INCOME SHIFTING Individuals possess entrepreneurial skill s ~H(s) over [ s, s]. 1 As worker earn w and pay PIT t 2 As small business entrepreneur earn π = s A and pay t as self-employed... but can also retain - shift - income and pay SBT τ < t 9 / 28

15 MODEL - OPTIMAL INCOME SHIFTING Individuals possess entrepreneurial skill s ~H(s) over [ s, s]. 1 As worker earn w and pay PIT t 2 As small business entrepreneur earn π = s A and pay t as self-employed... but can also retain - shift - income and pay SBT τ < t Entrepreneur s profit shifting problem: max θ }{{} s A t [s A θ] τθ θ2 2π =π... gives optimal amount of taxable income retained (shifted): θ (s, t,τ) = (t τ)s A 9 / 28

16 MODEL - OPTIMAL INCOME SHIFTING Individuals possess entrepreneurial skill s ~H(s) over [ s, s]. 1 As worker earn w and pay PIT t 2 As small business entrepreneur earn π = s A and pay t as self-employed... but can also retain - shift - income and pay SBT τ < t Entrepreneur s profit shifting problem: max θ }{{} s A t [s A θ] τθ θ2 2π =π... gives optimal amount of taxable income retained (shifted): θ (s, t,τ) = (t τ)s A Follows naturally that θ τ < 0 and θ s > 0 9 / 28

17 MODEL - OPTIMAL INCOME SHIFTING Individuals possess entrepreneurial skill s ~H(s) over [ s, s]. 1 As worker earn w and pay PIT t 2 As small business entrepreneur earn π = s A and pay t as self-employed... but can also retain - shift - income and pay SBT τ < t Entrepreneur s profit shifting problem: max θ }{{} s A t [s A θ] τθ θ2 2π =π... gives optimal amount of taxable income retained (shifted): θ (s, t,τ) = (t τ)s A Follows naturally that θ τ < 0 and θ s > 0 Could specify π(s,l) = s f (l) wl and f (l) = Al α 9 / 28

18 MODEL - ENDOGENOUS OCCUPATIONAL CHOICE Risk-neutral individuals occupational choice depends on tax rates. s - cutoff skill of the marginal individual who is indifferent between becoming a small business entrepreneur or a worker. 10 / 28

19 MODEL - ENDOGENOUS OCCUPATIONAL CHOICE Risk-neutral individuals occupational choice depends on tax rates. s - cutoff skill of the marginal individual who is indifferent between becoming a small business entrepreneur or a worker. The necessary condition: ) (t τ)2 (1 t + s A = (1 t)w 2 }{{}}{{} net wage income net-of-tax-shifting profit income 10 / 28

20 MODEL - ENDOGENOUS OCCUPATIONAL CHOICE Risk-neutral individuals occupational choice depends on tax rates. s - cutoff skill of the marginal individual who is indifferent between becoming a small business entrepreneur or a worker. The necessary condition: ) (t τ)2 (1 t + s A = (1 t)w 2 }{{}}{{} net wage income net-of-tax-shifting profit income ( w ) ( ) Solving for s 2(1 t) (t,τ) = A 2(1 t) + (t τ) 2 10 / 28

21 MODEL - ENDOGENOUS OCCUPATIONAL CHOICE Risk-neutral individuals occupational choice depends on tax rates. s - cutoff skill of the marginal individual who is indifferent between becoming a small business entrepreneur or a worker. The necessary condition: ) (t τ)2 (1 t + s A = (1 t)w 2 }{{}}{{} net wage income net-of-tax-shifting profit income ( w ) ( ) Solving for s 2(1 t) (t,τ) = A 2(1 t) + (t τ) 2 s τ > 0 As τ income shifting value and more i choose to become small business entrepreneurs... but also lowers average skill. 10 / 28

22 MODEL - AGGREGATES Aggregate small business profit income... s s Π(s (t,τ)) = π(s)h(s)ds = A s sh(s)ds s... shrinks when fewer people become entrepreneurs: Π s < 0 11 / 28

23 MODEL - AGGREGATES Aggregate small business profit income... s s Π(s (t,τ)) = π(s)h(s)ds = A s sh(s)ds s... shrinks when fewer people become entrepreneurs: Π s < 0 Aggregate small business income retained as profit... Θ ( s (t,τ) ) = (t τ) Π(s (t,τ))... increases in the difference between t and τ Important factor for why SB would want to lobby for a τ. 11 / 28

24 GOVERNMENT S OPTIMAL SMALL BUSINESS TAX RATE τ Government collects revenue from: workers: t w H(s ) small businesses: t [Π(s (t,τ)) Θ(s (t,τ))] + τθ(s (t,τ)) 12 / 28

25 GOVERNMENT S OPTIMAL SMALL BUSINESS TAX RATE τ Government collects revenue from: workers: t w H(s ) small businesses: t [Π(s (t,τ)) Θ(s (t,τ))] + τθ(s (t,τ)) Government s objective is a weighted revenue sum: max τ G(t,τ) = ( t (t τ) 2) Π(s (t,τ)) + λt w H(s ) λ > 0 is the political weight attached to revenue collected from workers 12 / 28

26 GOVERNMENT S OPTIMAL SMALL BUSINESS TAX RATE τ Government collects revenue from: workers: t w H(s ) small businesses: t [Π(s (t,τ)) Θ(s (t,τ))] + τθ(s (t,τ)) Government s objective is a weighted revenue sum: max τ G(t,τ) = ( t (t τ) 2) Π(s (t,τ)) + λt w H(s ) λ > 0 is the political weight attached to revenue collected from workers Not simultaneous decision on t and τ... take t as predetermined 12 / 28

27 GOVERNMENT S OPTIMAL SMALL BUSINESS TAX RATE τ Government collects revenue from: workers: t w H(s ) small businesses: t [Π(s (t,τ)) Θ(s (t,τ))] + τθ(s (t,τ)) Government s objective is a weighted revenue sum: max τ G(t,τ) = ( t (t τ) 2) Π(s (t,τ)) + λt w H(s ) λ > 0 is the political weight attached to revenue collected from workers Not simultaneous decision on t and τ... take t as predetermined Given the restriction t > τ 0, interior solution FOC is: 2(t τ )Π(s ) + ( t (t τ ) 2) Π s s τ + λt wh(s ) s τ = 0 for small λ more business friendly Government τ and (t τ) 12 / 28

28 SMALL BUSINESS BUYING A LOWER τ BY LOBBYING Small businesses might try to obtain an even lower τ through lobbying. simple Principal Agent setup max(1 T (t,τ))π(s (t,τ)) c subject to G(t,τ) + c G (t,τ ) τ 13 / 28

29 SMALL BUSINESS BUYING A LOWER τ BY LOBBYING Small businesses might try to obtain an even lower τ through lobbying. simple Principal Agent setup max(1 T (t,τ))π(s (t,τ)) c subject to G(t,τ) + c G (t,τ ) τ offer take-it-or-leave-it contribution c to obtain lower τ. FOC is ( t τ ) Π(s ) + [1 t + (t τ) 2 ] Π s 2 s τ + + 2(t τ)π(s ) + ( t (t τ) 2) Π s s τ + λt wh(s ) s τ Evaluate at τ = τ, the second line is = 0 and FOC < 0 now τ < τ 13 / 28

30 SMALL BUSINESS BUYING A LOWER τ BY LOBBYING Small businesses might try to obtain an even lower τ through lobbying. simple Principal Agent setup max(1 T (t,τ))π(s (t,τ)) c subject to G(t,τ) + c G (t,τ ) τ offer take-it-or-leave-it contribution c to obtain lower τ. FOC is ( t τ ) Π(s ) + [1 t + (t τ) 2 ] Π s 2 s τ + + 2(t τ)π(s ) + ( t (t τ) 2) Π s s τ + λt wh(s ) s τ Evaluate at τ = τ, the second line is = 0 and FOC < 0 now τ < τ Proposition 1 ( EMPIRICAL ) τ set under the influence of a small business lobby is lower than the revenue maximizing rate τ... regardless of political preference. 13 / 28

31 DETERMINANTS OF PROVINCIAL SMALL BUSINESS TAX RATE, Not the main empirical focus, just a brief evaluation of Prop 1 Ideally, test Prop 1 as Chirinko & Wilson [2010] analysis no political contribution or direct lobbying expenditure data. Instead, I create an indicator for years that CFIB was actively present 13 / 28

32 DETERMINANTS OF PROVINCIAL SMALL BUSINESS TAX RATE, Not the main empirical focus, just a brief evaluation of Prop 1 Ideally, test Prop 1 as Chirinko & Wilson [2010] analysis no political contribution or direct lobbying expenditure data. Instead, I create an indicator for years that CFIB was actively present τ r,s = α 1 LD r,s + Γ r,s γ + Σ r,s σ + θ r + ɛ r,s Γ r,s - time-varying provincial political control variables 1 Political Ideology: party score from 1 (right-wing) to -1 (left-wing) IDEOLOGY SCALE and IDEOLOGY VALUES 2 Election year dummy (could be lagged) 3 Herfindahl-Hirschman Index (HHI) of political competition 13 / 28

33 DETERMINANTS OF PROVINCIAL SMALL BUSINESS TAX RATE, Not the main empirical focus, just a brief evaluation of Prop 1 Ideally, test Prop 1 as Chirinko & Wilson [2010] analysis no political contribution or direct lobbying expenditure data. Instead, I create an indicator for years that CFIB was actively present τ r,s = α 1 LD r,s + Γ r,s γ + Σ r,s σ + θ r + ɛ r,s Γ r,s - time-varying provincial political control variables 1 Political Ideology: party score from 1 (right-wing) to -1 (left-wing) IDEOLOGY SCALE and IDEOLOGY VALUES 2 Election year dummy (could be lagged) 3 Herfindahl-Hirschman Index (HHI) of political competition Σ r,s vector of provincial economic controls 1 Real provincial GDP (deflated by provincial CPI, 2013 $) 2 Average regional small business tax (Atlantic, Central, West) 13 / 28

34 PROVINCIAL SMALL BUSINESS TAX DETERMINANTS, BACK Provincial SBIT Dependent variable: Combined SBIT (1) (2) (3) (4) Lobby Dummy (.165) (.801) (.347) (.958) p = p =.020 p = p =.008 Political Ideology (1.460) (.743) (1.910) (1.120) p =.552 p =.111 p =.548 p =.128 Election Dummy (.174) (.224) p =.956 p =.954 Herfindahl Hirschman Index (3.410) (3.850) p =.347 p =.100 Provincial Real GDP (.00001) (.00001) p =.522 p =.325 Regional SBIT (.178) (.242) p =.024 p =.029 Observations Adjusted R F Statistic Note: p<0.1; p<0.05; p<0.01 Reported in brackets are cluster-robust standard errors, clustered by province. 13 / 28

35 SMALL-BUSINESS TAXATION AND INCOME INEQUALITY Does lower τ increase income concentration? 14 / 28

36 SMALL-BUSINESS TAXATION AND INCOME INEQUALITY Does lower τ increase income concentration? Define R(p,τ; t) as the share of top p% income in total business income R(p,τ; t) = (1 T ) s s p π(s)h(s)d(s) (1 T ) s s π(s)h(s)d(s) 14 / 28

37 SMALL-BUSINESS TAXATION AND INCOME INEQUALITY Does lower τ increase income concentration? Define R(p,τ; t) as the share of top p% income in total business income R(p,τ; t) = (1 T ) s s p π(s)h(s)d(s) (1 T ) s s π(s)h(s)d(s) Primary interest: Does R(p,τ; t) increase when τ decreases? Need the functional distribution of s and the upper bound s. 14 / 28

38 SMALL-BUSINESS TAXATION AND INCOME INEQUALITY Does lower τ increase income concentration? Define R(p,τ; t) as the share of top p% income in total business income R(p,τ; t) = (1 T ) s s p π(s)h(s)d(s) (1 T ) s s π(s)h(s)d(s) Primary interest: Does R(p,τ; t) increase when τ decreases? Need the functional distribution of s and the upper bound s. 1 Untruncated Pareto distribution 2 Truncated Pareto distribution ( ) 1 z H(s) = 1 and h(s) = z, shape parameter z > 1 s s1+z 14 / 28

39 (1) UNTRUNCATED PARETO DISTRIBUTION, s = Aggregate after-tax-after-shifting profit is (1 T )Π(s (1 T )Az ) = s 1 z z 1 for s = {s, s p }. To derive income ratio R(p,τ; t) need to find the value of cutoff s p 15 / 28

40 (1) UNTRUNCATED PARETO DISTRIBUTION, s = Aggregate after-tax-after-shifting profit is (1 T )Π(s (1 T )Az ) = s 1 z z 1 for s = {s, s p }. To derive income ratio R(p,τ; t) need to find the value of cutoff s p s p (t,τ) = s (t,τ) p 1/z DERIVATION With this we have: R(p,τ; t) = (1 T )Π(s p(t,τ)) (1 T )Π(s (t,τ)) = s1 z p z 1 = p z s 1 z 15 / 28

41 (1) UNTRUNCATED PARETO DISTRIBUTION, s = Aggregate after-tax-after-shifting profit is (1 T )Π(s (1 T )Az ) = s 1 z z 1 for s = {s, s p }. To derive income ratio R(p,τ; t) need to find the value of cutoff s p s p (t,τ) = s (t,τ) p 1/z DERIVATION With this we have: Proposition 2 R(p,τ; t) = (1 T )Π(s p(t,τ)) (1 T )Π(s (t,τ)) = s1 z p z 1 = p z s 1 z When the entrepreneurial skill distribution H(s) is an unbounded Pareto distribution, the income concentration is independent of the business tax τ. 15 / 28

42 (2) TRUNCATED PARETO DISTRIBUTION, s = b More appropriate to evaluate the effect of τ on R(p,τ; t) using a skill distribution with a finite upper bound 16 / 28

43 (2) TRUNCATED PARETO DISTRIBUTION, s = b More appropriate to evaluate the effect of τ on R(p,τ; t) using a skill distribution with a finite upper bound Pareto CDF and PDF, truncated from the right, with support [1,b] : H b (s) := H(s) H(b) and h b (s) := h(s) H(b) 16 / 28

44 (2) TRUNCATED PARETO DISTRIBUTION, s = b More appropriate to evaluate the effect of τ on R(p,τ; t) using a skill distribution with a finite upper bound Pareto CDF and PDF, truncated from the right, with support [1,b] : H b (s) := H(s) H(b) and h b (s) := h(s) H(b) Then the aggregate after-tax profit income is: (1 T )Π(s (1 T )Az ( ) = s 1 z b 1 z) (z 1)H(b) for s = {s, s p }. DERIVATION The top p% income share with the truncated skill distribution takes the form R(p,τ; t) = s1 z p (t,τ) b 1 z s 1 z...which is not a constant! (t,τ) b1 z 16 / 28

45 (2) TRUNCATED PARETO DISTRIBUTION, s = b Proposition 3 ( PROOF ) When the entrepreneurial skill is Pareto distributed and truncated from the right, income inequality is negatively related to the SBT τ. R(p,τ; t) (1 z) s = τ } s {{ z } τ b1 z } {{ } + where z > 1, b > s > 0, and s τ > 0. [ pb + (1 p)s ] 1 [pb z + (1 p)s z ] }{{ 1/z < 0 } + 17 / 28

46 SMALL-BUSINESS TAXATION AND INCOME INEQUALITY R(p,τ; t) is a measure function, where R(0,τ; t) = 0, R(1,τ; t) = 1, For τ < τ we have R(p,τ ) > R(p,τ) graphically 18 / 28

47 SMALL-BUSINESS TAXATION AND INCOME INEQUALITY R(p,τ; t) is a measure function, where R(0,τ; t) = 0, R(1,τ; t) = 1, For τ < τ we have R(p,τ ) > R(p,τ) graphically 1 R(p, τ ) R(p, τ) p R(p,τ ) is first-order stochastically dominated by R(p,τ) Top percentile captures a higher income share under a lower τ 18 / 28

48 EMPIRICAL ANALYSIS OF PROPOSITION 3 Proposition 3 tells us that decreasing τ is not income inequality neutral. Expect that the top income shares increase as τ decreases. Explore theoretical implications with panel data of 10 Canadian provinces. 19 / 28

49 EMPIRICAL ANALYSIS OF PROPOSITION 3 Proposition 3 tells us that decreasing τ is not income inequality neutral. Expect that the top income shares increase as τ decreases. Explore theoretical implications with panel data of 10 Canadian provinces. Lack of precise measures for top income shares among different income sources and occupations... have to rely on public, aggregate top shares. Nevertheless, estimation still consistent b/c CCPC income augments top income p% while maintaining the pattern of fluctuation. Data Sources: Longitudinal Administrative Databank, CANSIM ( ) and the Canadian Tax Foundation. Summary Statistics 19 / 28

50 17 16 PROVINCIAL TOP 1% INCOME SHARES, % AB BC MB NB NL NS ON PE QC SK Total After Tax Income including Capital Gains. Test results sensitivity to the time period and income concept. 20 / 28

51 COMBINED SMALL BUSINESS TAX RATES, % AB BC MB NB NL NS ON PE QC SK Combined = Provincial + Federal statutory rates. 21 / 28

52 PREVIEW OF THE MAIN EMPIRICAL RELATIONSHIP PROVINCIAL 1% INCOME SHARE VS. SMALL BUSINESS TAX RATES AB BC MB NB NL NS ON PE QC SK Top 1% Income Share, ATI+CG AB BC MB NB NL NS ON PE QC SK Small-business income tax rates (%) 22 / 28

53 SPECIFICATION I S(p) r s = β 1 τ r s + β 2 y r s + β 3 X r s + θ r + µ t + ɛ r s Start with 1988, after major tax reform (260 obs.) I S(p) : Primarily Top 1%, but check 0.1% also Preferred income definition: ATI + CG 1 Use as complete income definition as possible when studying inequality 2 Also the definition used by Wolfson, Veall, et al. [2016] τ r s : combined federal and provincial tax rate for small businesses 23 / 28

54 SPECIFICATION I S(p) r s = β 1 τ r s + β 2 y r s + β 3 X r s + θ r + µ t + ɛ r s Start with 1988, after major tax reform (260 obs.) I S(p) : Primarily Top 1%, but check 0.1% also Preferred income definition: ATI + CG 1 Use as complete income definition as possible when studying inequality 2 Also the definition used by Wolfson, Veall, et al. [2016] τ r s : combined federal and provincial tax rate for small businesses y r s : provincial income control, real GDP per capita (2013$) U.S. top 1% control for global income inequality trend Saez & Veall [2005] show top shares in Canada follow U.S. ones closely θ r, µ t : province and time fixed effects All regressions weighted by provincial population (WLS) Also natural log to highlight the responsiveness of top shares to τ 23 / 28

55 CLUSTER-ROBUST VARIANCE ESTIMATION Issue in Canadian context: small number of clusters. CRVE performs poorly when number of clusters is small or even moderate (Donald & Lang [2007], Cameron & Miller [2015]) 1 CR errors are downward biased and need some correction 2 Even after bias correction, t-tests may over-reject in hypothesis tests 24 / 28

56 CLUSTER-ROBUST VARIANCE ESTIMATION Issue in Canadian context: small number of clusters. CRVE performs poorly when number of clusters is small or even moderate (Donald & Lang [2007], Cameron & Miller [2015]) 1 CR errors are downward biased and need some correction 2 Even after bias correction, t-tests may over-reject in hypothesis tests I rely on the approach of Bell & McCaffrey [2002]. 1 Bias-reducing correction; extension of HC2 var estimator to clustering 2 Suggest degrees-of-freedom correction based on Satterthwaite approx. Imbens & Kolesar [2016] advise to use these modifications more widely 24 / 28

57 CLUSTER-ROBUST VARIANCE ESTIMATION Issue in Canadian context: small number of clusters. CRVE performs poorly when number of clusters is small or even moderate (Donald & Lang [2007], Cameron & Miller [2015]) 1 CR errors are downward biased and need some correction 2 Even after bias correction, t-tests may over-reject in hypothesis tests I rely on the approach of Bell & McCaffrey [2002]. 1 Bias-reducing correction; extension of HC2 var estimator to clustering 2 Suggest degrees-of-freedom correction based on Satterthwaite approx. Imbens & Kolesar [2016] advise to use these modifications more widely Pustejovsky & Tipton [2016] generalization and extension of Bell & McCaffrey [2002] method. 24 / 28

58 TABLE 3 REGRESSION RESULTS, Dependent variable: Income Share of Top 1% (ATI + CG) (1) (2) (3) (4) (5) (6) (7) SBT (combined) (.004) (.003) (.007) (.007) (.014) (.005) (.028) p = p =.0005 p =.002 p =.002 p =.048 p =.008 p =.044 US Top 1%.273 (.008) p = Real GDP/capita (.00002) (.00002) (.00002) (.00002) p =.100 p =.080 p =.128 p =.059 Regional SBT.046 (.020) p =.105 Top PIT (combined) (.014) (.014) p =.002 p =.002 Year effect? No No Yes Yes Yes Yes Yes Province effect? Yes Yes Yes Yes Yes Yes Yes Province trend? No No No No No No Yes Adjusted R F Statistic Observations Note: p<0.1; p<0.05; p<0.01 Reported in parentheses are robust standard errors, clusterd by province. All regressions are weighted by the provincial population. 25 / 28

59 ROBUSTNESS ACROSS TIME PERIOD Top 1% (ATI + CG) Dependent variable: Top 0.1% (ATI + CG) (1) (2) (3) (4) (5) (6) SBT (combined) (.013) (.014) (.011) (.004) (.004) (.003) p =.081 p =.109 p =.212 p =.076 p =.084 p =.086 Real GDP/capita (.00002) (.00002) (.00001) ( ) ( ) ( ) p =.131 p =.154 p =.004 p =.008 p =.002 p =.005 Top PIT (combined) (.018) (.011) (.005) (.001) p =.072 p =.005 p =.171 p =.012 Year effect? Yes Yes Yes Yes Yes Yes Province effect? Yes Yes Yes Yes Yes Yes Province trend? No No Yes No No Yes Adjusted R F Statistic Observations Note: p<0.1; p<0.05; p<0.01 In the last three columns, income shares data for NL, PEI, NB are missing. Reported in parentheses are robust standard errors, clusterd by province. All regressions are weighted by the provincial population. 26 / 28

60 ROBUSTNESS FOR INCOME DEFINITIONS, Dependent variable: Top 1% (Total Income + CG) Top 1% (Total Income) (1) (2) (3) (4) (5) (6) SBT (combined) (.007) (.006) (.029) (.006) (.005) (.023) p =.016 p =.832 p =.062 p =.416 p =.022 p =.056 Real GDP/capita (.00002) (.00002) (.00002) (.00002) (.00002) (.00002) p =.125 p =.185 p =.058 p =.139 p =.223 p =.054 Top PIT (combined) (.015) (.014) (.017) (.010) p =.002 p =.002 p =.003 p =.002 Year dummies? Yes Yes Yes Yes Yes Yes Province dummies? Yes Yes Yes Yes Yes Yes Province trend? No No Yes No No Yes Adjusted R F Statistic Observations Note: p<0.1; p<0.05; p<0.01 Reported in parentheses are robust standard errors, clusterd by province. All regressions are weighted by the provincial population. 27 / 28

61 CONCLUSION Small businesses are important contributors to economic activity. employ 70.5% total private labour force; 30% of provincial GDP; responsible for 87% of net employment change yoy But also used as vehicles for significant tax avoidance by high incomes. 61% of Top 1%, 77% of Top 0.1% and 85% of Top 0.01% own CCPCs 28 / 28

62 CONCLUSION Small businesses are important contributors to economic activity. employ 70.5% total private labour force; 30% of provincial GDP; responsible for 87% of net employment change yoy But also used as vehicles for significant tax avoidance by high incomes. 61% of Top 1%, 77% of Top 0.1% and 85% of Top 0.01% own CCPCs Important contributors to overall income inequality developments and small business tax policy has a significant role 28 / 28

63 CONCLUSION Small businesses are important contributors to economic activity. employ 70.5% total private labour force; 30% of provincial GDP; responsible for 87% of net employment change yoy But also used as vehicles for significant tax avoidance by high incomes. 61% of Top 1%, 77% of Top 0.1% and 85% of Top 0.01% own CCPCs Important contributors to overall income inequality developments and small business tax policy has a significant role 1 Theoretical framework and mechanism that highlights the role of SBT policy in widening income inequality. 2 Show top p% capture higher share of income when SBT is lowered contributes to overall widening of income inequality. 3 Show how small business can obtain a lower tax rate by lobbying. 28 / 28

64 CONCLUSION Small businesses are important contributors to economic activity. employ 70.5% total private labour force; 30% of provincial GDP; responsible for 87% of net employment change yoy But also used as vehicles for significant tax avoidance by high incomes. 61% of Top 1%, 77% of Top 0.1% and 85% of Top 0.01% own CCPCs Important contributors to overall income inequality developments and small business tax policy has a significant role 1 Theoretical framework and mechanism that highlights the role of SBT policy in widening income inequality. 2 Show top p% capture higher share of income when SBT is lowered contributes to overall widening of income inequality. 3 Show how small business can obtain a lower tax rate by lobbying. FUTURE EXTENSIONS AND IMPROVEMENTS More precise estimate of the SBT effect on top income shares and develop a better specified model of joint tax rate determination. 28 / 28

65 Appendix PROVINCIAL PARTY IDEOLOGY, , BACK Instead of dummies identifying ideology, more refined coding PQ SP UN Social Credit WP ADQ NDP Liberal Conservative [Political Ideology] r s = p i ps S pr s p S pr s, where S pr s is the number of seats party p has in province r in year s. Ideological position of parties, and in turn provincial legislatures, changes over time as different factions are in power. (Bjornskov and Potrafke [2012]) 28 / 28

66 Appendix PROVINCIAL PARTY IDEOLOGY, , BACK PROVINCIAL PARLIAMENTS POLITICAL IDEOLOGY AB BC MB NB NL Index NS ON PE QC SK / 28

67 Appendix DERIVATION OF s p (t,τ) WITH UNBOUNDED PARETO DISTRIBUTION BACK For a certain percentile p, the value of s p is derived by solving: The equation above becomes Pr(s s s p ) Pr(s s = 1 p ) H(s p ) H(s ) 1 H(s = ) s p = 1 s 1 z sp z 1 s z s z s s z p p 1/z = 1 p = p 28 / 28

68 Appendix DERIVATION OF s p (t,τ) WITH BOUNDED PARETO DISTRIBUTION BACK For a certain percentile p, the value of s p is derived by solving: H b (s p ) H b (s ) 1 H b (s = 1 p where bounded Pareto CDF: H b (s) = H(s) ) H(b) H(s p ) H(s ) H(b) H(s ) = 1 s 1 z sp z 1 s z 1 b z b z (s z p s z ) s z p(b z s z ) = 1 p = 1 p b z s z = sp z [ pb z + (1 p)s z] s p = bs [ pb z + (1 p)s z] 1/z 28 / 28

69 Appendix PROPOSITION 3 PROOF The final bracket is positive when: 1 > pb + (1 p)s [ pb z + (1 p)s z] 1/z [ pb z + (1 p)s z] 1/z > pb + (1 p)s pb z + (1 p)s z > [ pb + (1 p)s ] z 28 / 28

70 Appendix PROPOSITION 3 PROOF The final bracket is positive when: 1 > pb + (1 p)s [ pb z + (1 p)s z] 1/z [ pb z + (1 p)s z] 1/z > pb + (1 p)s pb z + (1 p)s z > [ pb + (1 p)s ] z 28 / 28

71 Appendix PROPOSITION 3 PROOF The final bracket is positive when: 1 > pb + (1 p)s [ pb z + (1 p)s z] 1/z [ pb z + (1 p)s z] 1/z > pb + (1 p)s pb z + (1 p)s z > [ pb + (1 p)s ] z Strictly convex function for any z > 1. Then for p (0,1) and a pair (b, s ) such that b s, this is true by the definition of a strictly convex function. BACK 28 / 28

72 Appendix DERIVATION OF s p (t,τ) WITH UNBOUNDED PARETO DISTRIBUTION BACK Statistic N Mean St. Dev. Min Median Max Small Business Tax (provincial) Small Business Tax (combined) Top 1% Income Share Top 0.1% Income Share Provincial GDP/capita, $ ,700 12,405 24,641 40,479 87,282 U.S. Top 1% Income Share Top 1% Income Share, men Top 1% Income Share, women Top 0.1% Income Share, men Top 0.1% Income Share, women Note: All data at yearly frequency. Income is defined as after-tax including capital gains. Income shares separated by sex are not available for all provinces. 28 / 28