Employment Structure and the Rise of the Modern Tax System

Size: px

Start display at page:

Download "Employment Structure and the Rise of the Modern Tax System"

Paul Ford
6 years ago
Views:

1 Employment Structure and the Rise of the Modern Tax System Anders Jensen Harvard Kennedy School Yale - December 5th, 2016

2 Why does tax capacity grow over development? What this paper does

3 What this paper does Show how transition from self-employment to employee-jobs over development affects capacity to raise income tax New channel employee-share exemption threshold tax take I provide micro evidence on channel over long run of development Novel stylized facts Quasi-experimental evidence Unexploited place based program in US states

4 Descriptive evidence: new stylized facts India [$1034 pc] China [$1950 pc] Income Deciles Income Deciles Employee Self employed Employee Self employed Mexico [$7834 pc] US [$42000 pc] Income Deciles Income Deciles Employee Self employed Employee Self employed

5 Descriptive evidence: new stylized facts India [$1034 pc] China [$1950 pc] Income Deciles Income Deciles Employee Self employed Employee Self employed Mexico [$7834 pc] US [$42000 pc] Income Deciles Income Deciles Employee Self employed Employee Self employed

6 Quasi experimental evidence: US development program Use historical U.S. development program - Industrial Development Bonds (IDB) - to instrument for employee share Exploit within-state variation in court upholding of IDB program IDB implementation required House vote and Court upholding Treatment: State Court upholding event Control: State House vote event Relevance of US states for taxation-development Graph

7 Main contribution Previous evidence Micro evidence - lacks extension to tax capacity over development [Kleven et al., 2011] Macro evidence - lacks identified empirical channels [Besley & Persson, 2011; Kleven, Kreiner & Saez, 2015] I provide well-identified evidence going from employment structure via tax policy to tax revenue

8 Related literatures Recent literature on taxation and development Gordon and Li (2009), Besley and Persson (2011), Cage and Gadenne (2015), Kumler et al. (2015), Pomeranz (2015), Kleven, Kreiner and Saez (2015) Employment structure, informality over long run of development Gollin, Waugh and Lagakos (2013), LaPorta and Shleifer (2014), McNaig and Pavcnik (2014), Comin et al. (2015) US states as taxation-development laboratory Gillitzer (2013), Liu (2014), Dincecco and Troiano (2015) Place based local development programs Glaeser (2001), Story et al. (2013), Kline and Moretti (2014)

9 Outline 1 Stylized facts: employee-share and tax base over development 2 Quasi experimental evidence: employee-share impact on tax base and tax take 3 Model

10 Outline 1 Stylized facts: employee-share and tax base over development 2 Quasi experimental evidence: employee-share impact on tax base and tax take 3 Model

11 3 new stylized facts over development 1 Change in employee share distribution Within country employee share increases through income distribution Profile shifts leftward over development 2 Increase in size of income tax base Threshold moves progressively down, trailing employee share growth 3 No change to composition of income tax base Employee share above threshold constant at percent

12 Micro database: sources Household surveys: 90 countries and within US ( ) Search criteria: Continous measure of overall earned income Coverage of all types of work Sourced data directly with national statistical offices 83% of countries with per cap income $4, 000 Pre-1900 US

13 Micro database: methodology Code employee (E) if employment relation generates information trail relevant for tax enforcement Third party reporting, formal labor contracts Self-employed (SE) include domestic workers, non-registered family firm employees, casual daily wage laborers Captures ILO informal work category (2009) E-SE shares of non-agricultural employment in deciles of gross (pre-tax) income distribution Agriculture: evidence Go

14 3 new stylized facts over development 1 Change in employee share distribution Within country employee share increases through income distribution Profile shifts leftward over development 2 Increase in size of income tax base Threshold moves progressively down, trailing employee share growth 3 No change to composition of income tax base Employee share above threshold constant at percent

15 Start point: aggregate employee-share over development Employee share in total employment RWANDA INDIA CHINA INDONESIA MEXICO [Obs=90] Log real per capita income Country obs Local poly + 95% CI US Employee share in employment RWANDA INDIA US1870 US1940 US1930 US1920 SOUTH AFRICA US1910 DENMARK US1970 US1980 US1990 US2000 US1960 US2010 US Log real per capita income (Maddison data) Cross cty obs [n=67] Within cty obs: US [n=12]

16 Employee-share: cross-country Non agr employment share Non agr employment share $730 pc $13512 pc Non agr employment share Non agr employment share $3286 pc $53234 pc Employee Self employed

17 Employee-share: US & synthetic cross-country US Federal: Employee Cross cty synthetic: Employee US Federal: Self Emp Cross cty synthetic: Self Emp [India & Cambodia (ULHS); Brazil & Jordan (URHS); Lithuania & Slovakia (LLHS); Poland & Argentina (LRHS)]

18 3 new stylized facts over development 1 Change in employee share distribution Within country employee share increases through income distribution Profile shifts leftward over development 2 Increase in size of income tax base Threshold moves progressively down, trailing employee share growth 3 No change to composition of income tax base Employee share above threshold constant at percent

19 Fact #2 cross-country India [$1034 pc] China [$1950 pc] Income Deciles Income Deciles Employee Self employed Employee Self employed Mexico [$7834 pc] US [$42000 pc] Income Deciles Income Deciles Employee Self employed Employee Self employed

20 Fact #2 within country US Fed: 1870 US Fed: 1935 Income deciles Income deciles Employee Self employed Employee Self employed US Fed: 1950 US Fed: 1960 Income deciles Income deciles Employee Self employed Employee Self employed

21 Fact #2 cross and within country PIT base share in total employment RWANDA INDIA CHINA INDONESIA MEXICO [Obs=90] Log real per capita income Country obs Local poly + 95% CI US PIT base share in total employment RWANDA INDIA SOUTH AFRICA US1950 US1930 US1960 US1970 US1980 DENMARK US2000 US1990 US Log real per capita income (Maddison data) Cross cty obs [n=67] Within cty obs: US [n=12]

22 3 new stylized facts over development 1 Change in employee share distribution Within country employee share increases through income distribution Profile shifts leftward over development 2 Increase in size of income tax base Threshold moves progressively down, trailing employee share growth 3 No change to composition of income tax base Employee share above threshold constant at percent

23 Fact #3 cross and within country Employee share in PIT base RWANDA INDIA INDONESIA CHINA MEXICO [Obs=90] Log real per capita income Country obs Local poly + 95% CI US Employee share in PIT base RWANDA INDIA US1960 DENMARK US1970 US1980 US2010 SOUTH US1930 AFRICA US1950 US1990 US Log real per capita income (Maddison data) Cross cty obs [n=67] Within cty obs: US [n=12]

24 Stylized facts: robustness 1 Micro data external validity World Bank agriculture data-base Go 2 Different within country series Brazil Go 3 Confounding constant subsistence threshold Real value of threshold over development Go 4 Confounding distributional variation Sectoral: manuf + non-trade services Go Firm-size: prevalence of small firms Go

25 Stylized facts: implications and limitations Stylized facts consistent with proposed channel employee-share exemption threshold tax base Hold over full development path (across and within country) Suggest, but cannot conclude, a causal channel

26 Outline 1 Stylized facts: employee-share and tax base over development 2 Quasi experimental evidence: employee-share impact on tax base and tax take 3 Model

27 Quasi-experimental evidence: US development program Identify the impact of employee-share on the tax exemption threshold Require exogenous variation in employee share Historical U.S. development program to instrument for employee share U.S. states: compelling setting for taxation-development

28 State employee share and income tax threshold [ ] Share of decile employment Share of decile employment Income decile Income decile Employees Self employed Employees Self employed Share of decile employment Share of decile employment Income decile Income decile Employees Self employed Employees Self employed

29 Program details: Industrial Development Bonds (IDB) IDB targeted redevelopment counties (ARA, 1958) Surplus labor engaged in self-employed agriculture Credit constraints limiting plant expansions IDB constructed leasable plants financed by revenue bonds Led to employee share Constitution prohibits using public credit for private purpose Implementation required state legislature vote to enact IDB statute and court upholding of statute

30 Identification y st = β + α1 (Vote in) st + θ1 (Uphold) st + λx st + µ s + γ t + φ s t + ε st Identification assumption: conditional on vote in, timing of upholding is orthogonal to state-time unobservable determinants of employee-share Evidence for identifying assumption 1 Sharp on impact changes, break from stable pretrend vote-period 2 Timing explained by state-specific time-invariant chars. Hazard model

31 Within state IDB variation: legal decisions and issuance Colorado Delaware Georgia Illinois # issues # issues # issues # issues Years to/since vote in Years to/since vote in Years to/since vote in Years to/since vote in Iowa Maryland Minnesota Missouri # issues $ millions # issues $ millions # issues $ millions # issues $ millions Years to/since vote in Years to/since vote in Years to/since vote in Years to/since vote in New Hampshire North Dakota Ohio Pennsylvania # issues $ millions # issues $ millions # issues $ millions # issues $ millions Years to/since vote in Years to/since vote in Years to/since vote in Years to/since vote in South Carolina Vermont Virginia Washington # issues $ millions # issues $ millions # issues $ millions # issues $ millions Years to/since vote in Years to/since vote in Years to/since vote in Years to/since vote in

32 Determinants of lag between vote and upholding LHS=1(Upheld) Panel A: Politics 1(Civil law) (3.201)*** (5.212)*** (4.592)*** (4.815)*** (5.489)*** (3.884)*** (5.031)*** (4.921)*** 1(Hist rail default) (.105)*** (.334) (.120)*** (.116)*** (.449) (.108)*** (.215)* Covariate - - SupremeC Election Voting Population Neighbor RightWork appointed? dummies restrictions total uphold? laws? Signif 10 pct? - - N N N N N N Panel B: Economics 1(Civil law) (6.981)*** (3.905)*** (4.182)*** (5.134)*** (6.031)*** (4.370)*** (5.221)*** (5.988)*** 1(Hist rails) (.092)*** (.455) (.350) (.148)** (.078)*** (.208)* (.121)*** (.115)*** Covariate Redev LabForce Manuf Employee Income PIT PIT TotT counties growth growth share per cap base take take Signif 10 pct? N N N N N N N N Non-parametric Cox proportional hazard model, hazard rate reported

33 Exclusion restriction Objective: obtain causal estimate of employee share on tax threshold Instrument for employee share using IDB court upholding Exclusion restriction: IDB court upholding only affects changes in tax threshold location through changes in employee share Evidence for exclusion restriction 1 Full impact on employment and threshold materialize in the short run 2 No impact on determinants of threshold any non-employment economic outcome any tax revenue or instrument other than income tax revenue

34 Within-state event study: employee share Share employee indexed==1 at event year= Years to/since event Treatment: Upholding event Control: Vote in event

35 Within-state event study: employee share Share employee indexed==1 at event year= Years to/since event Treatment: Upholding event Control: Vote in event

36 Within-state event study: likelihood of threshold reform Empirical CDF # K reforms, indexed=1 at event year== Years to/since event Treatment: Upholding event Control: Vote in event

37 Within-state event study: size of tax base [Threshold/avg income] indexed==1 at event year= Years to/since event Treatment: Upholding event Control: Vote in event

38 Employment and earnings: different specifications Sample Full sample IDB states Cross-state Interval around reform (1) (2) (3) (4) (5) (6) E-share Avg inc E-share Avg inc E-share Avg inc 1(Vote) ( ) (423.2) ( ) ( ) 1(Uphold) *** *** *** ( ) (303.2) (.0062) ( ) ( ) (135.9) State FE, Year Fe x x x x x x State-year controls x x x x x x State linear trend x x x x State-year obs 3,384 3,384 2,664 2, Sample: , 48 states. S.E. clustered at state level.,, : significance at the 10%, 5%, 1%

39 Employment and earnings: full results Employment-share Average income Log(transfers) E SE Total E SE Non-E E-specific (1) (2) (3) (4) (5) (6) (7) 1(Vote) , ( ) ( ) (423.2) (317.4) (874.1) (0.0376) (0.0238) 1(Uphold) *** *** ** ( ) ( ) (303.2) (347.3) (672.8) (0.0220) (0.0219) State FE, Year FE x x x x x x x State linear trend x x x x x x x State-year controls x x x x x x x State-year obs 3,384 3,384 3,384 3,384 3,384 3,384 3,384 Sample: , 48 states. S.E. clustered at state level.,, : significance at the 10%, 5%, 1% y st = β + α1 (Vote in) st + θ1 (Upheld) st + λx st + µ s + γ t + φ s t + ε st

40 Employment and earnings: county level evidence Findings from county-level regressions [ ] Pre-post uphold versus pre-post vote in IDB versus non IDB county Federal classification of IDB counties (ARA, 1958) Findings: IDB impact employment-structure, only in IDB counties Absence of cross-county impacts between IDB and non-idb counties Absence of non-employment economic impacts in IDB counties Findings consistent with Federal characterization of IDB counties Existence of local labor surplus and private credit constraints Go

41 Employment and earnings: county level regressions E-share Agr-share Lg(E avg inc) Lg(rural pop) Lg(urban pop) Lg(value housing) (1) (2) (3) (4) (5) (6) 1(Vote) * ( ) ( ) (0.0234) (0.0172) (0.0248) (0.0353) 1(Uphold) ( ) ( ) (0.0223) (0.0174) (0.0310) (0.0268) 1(Vote)x1(IDB county) ( ) (0.0103) (0.0449) (0.0121) (0.0190) (0.0342) 1(Uphold)x1(IDB county) *** *** ( ) ( ) (0.0294) (0.0162) (0.0397) (0.0279) County FE, Year FE x x x x x x County linear trend x x x x x x County-year controls x x x x x x County-year obs 27,385 27,352 26,746 27,297 27,986 21,325 Sample: , all counties c in IDB states s. S.E. clustered at state level.,, : significance at the 10%, 5%, 1% y cst =β + α1 (Vote in) st + θ1 (Upheld) st + δ1 (Vote in) st 1(IDB county) c + π1 (Upheld) st 1(IDB county) c + λx cst + µ c + γ t + φ c t + ε cst

42 Exemption threshold K: reduced form and IV Reduced form 2nd stage IV [K/y] [Avg inc>k] [Avg inc<k] [K/y] [K/y] [K/y] (1) (2) (3) (4) (5) (6) 1(Vote) (0.191) (972.4) (160.9) 1(Uphold) ** (0.235) (902.9) (473.3) E-share ** ** ** (4.754) (12.45) (14.20) Instruments 1(Vote), 1(Uphold) 1(Uphold) 1st stage F-statistic P-value (0.0080) (0.0027) State FE, Year FE x x x x x x State linear trend x x x x x x State-year controls x x x x x x State-year obs 3,384 3,384 3,384 3,384 3,384 2,664 Sample: , 48 states. S.E. clustered at state level.,, : significance at the 10%, 5%, 1% y st = β + α1 (Vote in) st + θ1 (Upheld) st + λx st + µ s + γ t + φ s t + ε st

43 Exemption threshold K: exclusion restriction Growth: short-run (<10y) versus long-run (>10y) Confounding determinants of K Lg(K/y) Lg(E-share) Lg(E avg inc) 1(reform K) Income ineq [Hazard K] PIT rate 1(PIT WithH) TaxAdmin PolComp (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) 1(Vote) (0.190) ( ) (0.0142) (0.479) (0.0312) (0.387) ( ) (0.0877) (0.0926) (0.0107) 1(Uphold:0-10) *** *** ** (0.045) ( ) (0.0127) (0.168) (0.0283) (0.524) ( ) (0.0770) (0.0949) (0.0106) 1(Uphold: >10) ** * (0.172) ( ) (0.0263) (0.201) (0.0235) (0.306) ( ) (0.0600) (0.0760) (0.0109) State FE, Year FE x x x x x x x x x x State linear trend x x x x x x x x x x State-year controls x x x x x x x x x x State-year obs 3,384 3,347 3,431 3,394 2,911 2,478 3,456 2,592 3,024 3,456 Sample: , 48 states. S.E. clustered at state level.,, : significance at the 10%, 5%, 1% y st = β + α1 (Vote in) st + θ1 (Years since uphold 10) st + π1 (Years since uphold > 10) st + λx st + µ s + γ t + φ s t + ε st

44 Exemption threshold K: mechanisms [E-share income>k] E-share [K/y] (1) (2) (3) 1(Vote) (0.0145) (.0090) (0.230) 1(Vote)x1(EoI) (.0056) (0.151) 1(Uphold) *** ** (0.0128) (.0089) (0.291) 1(Uphold)x1(EoI).011** (.005) (0.250) State FE, Year FE x x x State linear trend x x x State-year controls x x x State-year obs 3,384 3,384 3,384 Sample: , 48 states. S.E. clustered at state level.,, denote significance at the 10%, 5%, 1% y cst =β + α1 (Vote in) st + θ1 (Upheld) st + δ1 (Vote in) st 1(EoI exists) st + π1 (Upheld) st 1(EoI exists) st + λx cst + µ c + γ t + φ c t + ε cst

45 Employment and threshold: robustness DiD specification and long time series Non-parametric significance level through permutation tests Go DiD small time window centered around reform Alternative controls Go Linear trends for structural determinants Non-parametric time path of civil law states

46 Tax revenue: reduced form and IV Reduced form 2nd stage IV PersIncT nonpit GSalesT SSalesT CIT LicenceT PropT PIT PIT PIT (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) 1(Vote) (0.121) (0.088) (0.079) (0.039) (0.022) (0.021) (0.007) 1(Uphold) 0.161** (0.078) (0.096) (0.073) (0.031) (0.026) (0.021) (0.037) E-share * 7.501** (1.182) (3.510) (3.295) Instruments 1(Vote), 1(Uphold) 1(Uphold) 1st stage F-statistic P-value (0.0080) (0.0027) State FE, Year FE x x x x x x x x x x State linear trend x x x x x x x x x x State-year controls x x x x x x x x x x State-year obs 3,384 3,384 3,384 3,384 3,384 3,384 3,384 3,384 3,384 2,664 Sample: , 48 states. S.E. clustered at state level.,, denote significance at the 10%, 5%, 1% y st = β + α1 (Vote in) st + θ1 (Upheld) st + λx st + µ s + γ t + φ s t + ε st

47 Outline 1 Stylized facts: employee-share and tax base over development 2 Quasi experimental evidence: employee-share impact on tax base and tax take 3 Model

48 Model: threshold K rule K rule: start at top, incremental decreases dk < 0 until K : MC(K )=MB(K ) Marginal benefit: dm = mechanical revenue gain > 0 Marginal costs Administrative: dc = dk c (ϕ K ) < 0 (Yitzhaki, 1979) ( ) Statutory employee-share>k Horizontal Equity: dh 1 Compliant 0 employee-share>k (Kopczuk, 2001) Efficiency: db = dk ε z,k ϕ K 0

49 Model: threshold K formula Rule delivers K formula K Main empirical prediction z = 1 [ [ ] Mech gain dc(ϕk ) 1 + Behavioral loss [ε z,k ϕ K ] 1] K ϕ K 0

50 Model: threshold K country vs K observed country PIT base share in total employment R 2 =.625 KENYA INDIA KENYA RWANDA INDIA CHINA CHINA INDONESIA INDONESIA RUSSIA RUSSIA MEXICO MEXICO SPAINUS ITALY US SPAIN ITALY Log real per capita income Actual country obs Predicted country obs Country obs: local poly Predicted obs: local poly

51 Conclusion Rise of modern tax system, driven by broad based PIT Micro evidence - lacks extension to tax capacity over development Macro evidence - lacks identified channels This paper: new PF-Dev research design to bridge micro and macro New channel: distributional employee share and tax threshold Descriptive evidence and quasi-experimental evidence Research design: potentially applicable to other PF-Dev questions Micro evidence: firm connectedness VAT enforcement Macro evidence: development firm connectedness

52 Appendix

53 Real value of threshold over development Back Exemption threshold in real US dollars [Obs=90] CHINA INDIA BANGLADESH RWANDA INDONESIA MEXICO DENMARK US Log real per capita income Country obs Local poly + 95% CI

54 Relevance of US states Personal income share in total tax Beta=.0558 (s.e.=.0012) Individual Income Tax/Total Tax Beta=.1758 (s.e.=.0025) Log per capita constant real income Bachas & Jensen (2015) Country year obs Linear Fit +95%CI Lg per capita constant real income State Year obs Linear Fit + 95% CI Back

55 Equal earnings pre-idb in IDB counties 1947 sample Earnings per Earnings per Difference [p-value] manufacturing job farmer (1) (2) (3) (4) IDB counties (10.28) (15.94) (12.70) [0.22] Non-IDB counties (14.21) (6.58) (12.81) [0.00]*** 1952 sample IDB counties (24.72) (51.81) (43.78) [0.77] Non-IDB counties (50.97) (17.98) (47.89) [0.00]*** Earnings per farmer constructed as [value of farm products sold + value of farm land] divided by number of farmers. Average wage-salary by manufacturing production worker.,, denote significance at the 10%, 5%, 1% level. Back

56 Robustness: regressions Back Specification Time path civil law Structural trends Exclude TVA states E-share [K/y] E-share [K/y] E-share [K/y] (1) (2) (3) (4) (5) (6) 1(Vote) ( ) (0.188) ( ) (0.191) ( ) (0.192) 1(Uphold) *** *** *** ** *** ** ( ) (0.227) ( ) (0.235) ( ) (0.180) State FE, Year FE x x x x x x State linear trend x x x x State-year controls x x x x x x State-year obs 3,384 3,384 3,384 3,384 3,096 3,096 Sample: , 48 states. S.E. clustered at state level.,, denote: at the 10%, 5%, 1%

57 Robustness: non-parametric permutation tests Back Empirical CDF p value= Estimated placebo coefficient 1(Uphold) Empirical CDF p value = Estimated placebo coefficient 1(Uphold)

58 Back Robustness: Brazil Non Ag Employee (LHS) Non Ag Selfemp (LHS)

59 Back Distributional sector-share: all countries $277 per cap $3286 per cap Non Agric Employment share Non Agric Employment share $13512 per cap $53234 per cap Non Agric Employment share Non Agric Employment share Manuf M + non trade Services

60 Distributional firm size-share: low-middle income countries $1422 pc $3286 pc Non Agr Employment share Non Agr Employment share Deciles of income distribution Deciles of income distribution $4638 pc $6945 pc Non Agr Employment share Non Agr Employment share Deciles of income distribution Deciles of income distribution Firm size <=10 Back

61 Distributional firm size-share: upper income countries $13512 pc $27596 pc Non Agr Employment share Non Agr Employment share Deciles of income distribution Deciles of income distribution $37369 pc $53234 pc Non Agr Employment share Non Agr Employment share Deciles of income distribution Deciles of income distribution Firm size <=10 Back

62 Distributional agriculture-share: low income countries $277 pc $730 pc Employment share Employment share Deciles of income distribution Deciles of income distribution $1422 $3286 Employment share Employment share Deciles of income distribution Deciles of income distribution Agric Employee within Agric Back

63 Distributional agriculture-share: middle income countries $4638 pc $6945 pc Employment share Employment share Deciles of income distribution Deciles of income distribution $13512 pc $27596 pc Employment share Employment share Deciles of income distribution Deciles of income distribution Agric Employee within Agric Back

64 Distributional agriculture-share: high income countries Employment share Employment share Deciles of income distribution Deciles of income distribution Agric Employee within Agric Agric Employee within Agric Back

65 Micro database vs World Bank: agriculture estimates Share aggregate employment Per capita income deciles Estimated Non Agr Est. Agriculture World Bank Agr Back

66 Model: τ formula At SWF optimum: dm + db + dc = 0 τ [ ( ω )] 1 τ = 1 g λ [ Mechanical gain Behavioural loss ] 1 [ε z,1 τ ϕ K ] In US setting: τ following dk < 0? Back ϕ K : no change [ ] Mechanical gain Ratio Behavioural loss : Social weight g ( ) ω λ :

Taxation and Development

Taxation and Development Henrik Kleven London School of Economics January 2017 1 / 51 What Separates PF-Devo From PF? PF-Devo is more than just studying taxation in developing countries Focus on tax enforcement