State Taxes and Spatial Misallocation

Transcription

1 State Taxes and Spatial Misallocation Pablo Fajgelbaum, Eduardo Morales, Juan Carlos Suárez Serrato, Owen Zidar UCLA, Princeton, Duke, Chicago Booth February 2016

2 Motivation Regional fiscal autonomy varies considerably across countries, e.g., France, Japan, United Kingdom: regional entities do not set tax policy Canada, Spain, taly: regions have large tax autonomy U.S. state tax autonomy is similar to the later group What are the aggregate consequences of regional di erences in taxes? Misallocation logic suggests potentially negative e ects F Restuccia and Rogerson (2008); Hsieh and Klenow (2009) This paper: quantify the aggregate e ects of dispersion in tax rates across U.S. states

3 Background on the U.S. State Tax System 3 Major Sources of Tax Revenue used by the States (4% of U.S. GDP) Sales tax (47% of total state tax revenue in 2012) more Personal income tax (35%) more Corporate income tax (5%) more State tax rates are heterogeneous percentiles of state tax-rate distributions: rates revenue F F F 7%-1.4% (sales) 9.2%-0% (personal income) 4.6%-0% (corporate) Changes in state tax revenue are strongly correlated with changes in direct expenditures E.g., education, public welfare, hospitals, highways, police, natural resources, parks and recreation.

4 This Paper 1 Quantitative Geography Model with U.S. State Tax System States with heterogeneous fundamentals F productivity, amenities, trade costs, factor shares, fixed factors, ownership rates Workers and firms sort across states according to idiosyncratic draws Firms are monopolistically competitive 3 major state taxes and federal transfers F 2 Estimation finance state spending which may be valued by workers and firms Elasticities of worker and firm location with respect to taxes F data on economic activity and 350 state tax changes over Fundamentals F match distribution of employment, wages, and trade in Main Counterfactual Movement to harmonized state tax system keeping government spending constant through inter-state transfers

5 Literature Misallocation Across firms: Restuccia and Rogerson (2008), Hsieh and Klenow (2009) Across cities: Brandt et al. (2013), Desmet and Rossi-Hansberg (2013), Moretti and Hsieh (2015) Rural-Urban gaps: Gollinetal.(2013),LagakosandWaugh(2013),Bryanand Morten (2015) Trade/Economic Geography Spatial equilibrium models: Roback(1982),Krugman(1991),Krugmanand Venables (1995), Helpman (1998), Tabuchi and Thisse (2002) Quantitative geography model: AllenandArkolakis(2014),Caliendoetal.(2014), Ramondo et al. (2015), Redding (2015), Gaubert (2015) Public Finance/Urban Fiscal competition: Flatters et al. (1973), Gordon (1983), Oates (1999), Keen and Konrad (2013), Ossa (2015) Factor Mobility and Policy Changes: Bartik(1991),Holmes(1998),Morettiand Wilson (2014), Suárez Serrato and Zidar (2015) General Equilibrium: Shoven et al. (1972), Ballard et al. (1985), Altig et al. (2001), Albouy (2009)

6 Model

7 Worker Utility and Location Fraction of workers who choose region n : vn L n = v State-specific appeal : where v n = u n Gn L W n Representative worker s utility: "W W,n 1 W (1 T n) wn,n P n 1 T n (1 ty n )(1 t y fed ) tw fed 1+t c n v X n v " W n! 1/"W

8 Firms Profits and Location Fraction of firms who choose region n: M n = n z 1 n! " F 1 where zn 1 = z 1 F,n n F G,n n Mn F Profits of firm with productivity z located in i: i (z) =max(1 t i ) NX n=1 x ni N X n=1 ni c i z q ni! where t i = t corp fed + t l i + P N n=1 tx n s ni Dispersion in sales apportionment leads to price distortion more mmobile capital owners in state n own a fraction b n of a portfolio that includes all firms and fixed factors

9 State Governments Government Spending=Tax Revenues+Federal Transfers P ng n = R n + T fed!st n State tax revenue Federal transfers R n = R corp n T fed!st n + R c n + R y n = nr n F F Running ln(p ng n)=ln Tn fed!st +ln(r nt)+" nt yields R 2 =0.97 Direct expenditures well approximated as a state-specific multiplier of tax revenue

10 Equilibrium n equilibrium, {L n, M n} N n=1, such that: fed Qn, Cn, n, Gn, Gn N n=1, {wn, rn}n n=1,and{pn}n n=1 Final-good producers optimize Workers make consumption and location decisions optimally Firms choose location, production and trade optimally Government budget is balanced in every location Local goods and factor markets (fixed factors and labor) clear in every location

11 Adjusted Fundamentals Taxes impact outcomes through: Adjusted fundamentals : A in = t in in, zn A / (1 t n) "F + F F PnG n GDP n W un A / (1 T n) 1 W PnG n u n GDP n F z 1 n F where PnGn GDP n is a function of taxes, parameters, and trade imbalances Trade imbalances (due to corporate taxes) mplementing counterfactuals with respect to taxes requires: mapping from changes in fundamentals to changes in outcomes (standard) mapping from changes in taxes to changes in adjusted fundamentals (specific to our model)

12 mpact of Tax Dispersion on Worker Welfare What is the e ect of eliminating dispersion in taxes {T n} on worker welfare v keeping {G n} constant? Assume F no trade frictions ( in =1),perfectsubstitutes( = 1), homogeneous firms (" F = 1), F no cross-state dispersion in income shares ( n = )orinpreferencesfor spending ( W,n = W ) Let Z n / z 0 n / n 1/ n H n u ng W n 1/(1 W ) and 1 W 1/" W + W W +(1 W ) Eliminating dispersion in{t n} while keeping spending constant increases (decreases) worker welfare if corr Zn, (1 T n) is low (high) enough ntuition Workers gain from dispersion in state-specific appeals, {v n} State-specific appeal depends on both taxes and fundamentals (through prices) High correlation between keep rates and fundamentals implies high dispersion Consider corr Zn, (1 T n) =0. Eliminating dispersion increases welfare if (1 W )(1 ) > 1/" W + W Holds for W,, W large, " W small

13 mpact of Tax Dispersion on Real ncome Eliminating dispersion in {T n} while keeping spending constant: may increase or decrease aggregate real income depending on the joint distribution of T n, u n,andg n increases it under perfect mobility (" W!1), no public goods ( W =0),andno dispersion in amenities (u n =0) ntuition F n this case, the model is the same as Hsieh and Klenow (2009) Real income maximized under no dispersion of MPL. But labor mobility is a ected by compensating di erentials, v = u ng W n ((1 T n)(mpl n)) 1 W

14 Key Forces Agglomeration Home market e ects due to trade costs (final consumption and intermediates): {" W," F,,{ n}} Returns to public spending: { F,n, W,n, F, W } Congestion Fixed factor used in production (land and structures): { n} Dispersed ownership of fixed factors: {b n} Heterogeneity in productivity, endowments, amenities, trade costs: {z n, H n, u n, in }

15 Data and Estimation

16 Data Economic Activity Number of workers and establishments by state (County Business Patterns) Total sales (Economic Census) Factor payments, value added (BEA) Trade flows (Commodity Flow Survey) Fiscal Policy Corporate tax rates and rules (Suárez Serrato and Zidar, 2014) ndividual income taxes (NBER TAXSM) Sales tax rates (Book of States) Government revenue and spending (Census of Governments)

17 Estimation Strategy Elasticities of firm and worker location estimated from data on economic activity and taxes from {" W, W,n } estimated from labor-supply equation {" F, F } estimated using firm-mobility equation Fundamentals, technologies, and ownership rates calibrated to (exactly) match data in 2007 {z n, H n, u n, in } chosen to match spending shares, sales shares, employment, and wages details { n, n} match input and labor shares details {b n} match trade deficits details Other parameters: Demand elasticity: = 4(Head et al.,2013) { W, F }2{{0, 0}, {1, 1}} (re-estimate in each case)

18 Estimation of {" W, W } Labor share in state n at time t: ln (L nt) = " W 1 W,n ln (1 T nt) wnt " W W,n Rnt + ln 1+ W " W W,n P nt 1+ W " W W,n P nt + L t + L n + L nt dentification assumption: E[ L nt L, L, Z L nt] =0 We use other-state taxes as instruments: Z L nt =(t c nt, t x nt, t y nt )

19 Estimation of {" W, W } Labor share in state n at time t: ln (L nt) =1.09 ln (1 T nt) wnt ln P nt Rnt P nt + L t + L n + L nt Case " W W W =0 W =1 W =0 W =1 Estimate W 1.39 (0.34) 2.01 (0.76) 0.22 (0.07) 0.22 (0.07) Estimate W,n = POL n [0.14, 0.18] [0.16, 0.17] Assume W =0 Assume F =0.05 = 1 N Assume W,n = Rn GDP n P R n n GDP n 1.04 (0.31) 1.15 (0.32) 1.25 (0.36) 1.04 (0.31) 1.22 (0.36) 1.4 (0.49)

20 Estimation of {" W, W } Labor share in state n at time t: ln (L nt) =1.09 ln (1 T nt) wnt ln P nt Rnt P nt + L t + L n + L nt Case " W W W =0 W =1 W =0 W =1 Estimate W 1.39 (0.34) 2.01 (0.76) 0.22 (0.07) 0.22 (0.07) Estimate W,n = POL n [0.21, 0.22] [0.21, 0.23] Assume W =0 Assume F =0.05 = 1 N Assume W,n = Rn GDP n P R n n GDP n 0.93 (0.28) 1.09 (0.31) 1.19 (0.35) 0.93 (0.28) 1.15 (0.35) 1.33 (0.45)

21 Estimation of {" F, F } Firm share in state n at time t: Rnt ln M nt =0.89 ln ((1 t n) MP nt)+0.14 ln P nt 2.69 ln c nt + M t + M n + M nt Case " F F F =0 F =1 F =0 F =1 Estimate F Assume F =0 Assume F = (0.33) 2.70 (0.61) 2.67 (0.32) 3.15 (0.77) 3.12 (0.45) 2.67 (0.32) 0.05 (0.06) 0.05 (0.06) details

22 Counterfactuals

23 E ect of 1pp Reduction in ncome Tax (G constant) Change in Own Rest of U.S. Keep Rate (1 T n) 1.12% 0% Employment 0.84% -0.02% (Pre-tax) Nominal Wage -0.43% 0.01% Firms 0.41% -0.01% Real GDP 0.52% -0.01% State E ect (v n) 0.44% 0.01% 1 T 0 n ln ˆLn =1.09 ln 1 T n,2007 {z } 1.14% ln ŵ n/ ˆP n {z } 0.32% 1.39 ln (ˆv) {z } 0.02%

24 1pp Reduction in California s ncome Tax (G constant) Figure : Employment Change (% points) CA = 0.84

25 mplementing Spending-Constant Counterfactuals Goal: assess the impact of tax dispersion keeping government spending constant n o Replace t y n,2007, tc n,2007, tn,2007, l tn,2007 x with (t y ) 0, (t c ) 0, t l 0, ( 0 t x ) s.t. 0 tn j = a j + b t j n,2007 for n =1,..,N, j = y, c, l, x and a j, b 0. fig. Keep government spending {G n} constant through an inter-state transfer system..e., for each b, find a j such that NX PnG 0 n,2007 = n=1 NX n=1 Rn 0 + T fed!st n Compute the change in aggregate real income and worker welfare, 0 bv = X n L n,2007 bv " W n 0! 1 " W

26 Eliminating Tax Dispersion R n 0 GDP n 0 Parametrization Constant Spending Variable Spending W,n F Welfare GDP Welfare GDP % 0.12% 0.98% 0.92% 0 + 1POL n % 0.12% 0.96% 0.93% R n GDP n % 0.12% 0.39% 0.94% of random n 0 6= n % 0.11% 0.38% 0.95% % 0.10% 0.16% 0.10% % 0.10% 0.48% -0.05% % 0.11% 0.44% 0.93%

27 Eliminating Tax Dispersion R n 0 GDP n 0 Parametrization Constant Spending Variable Spending W,n F Welfare GDP Welfare GDP % 0.12% 0.98% 0.92% 0 + 1POL n % 0.12% 0.96% 0.93% R n GDP n % 0.12% 0.39% 0.94% of random n 0 6= n % 0.11% 0.38% 0.95% % 0.10% 0.16% 0.10% % 0.10% 0.48% -0.05% % 0.11% 0.44% 0.93%

28 Allowing for Progressive ncome Taxes W,n F Welfare GDP A. State Progressive Only % 0.11% % 0.10% R n GDP n % 0.11% B. State and Federal Progressive % 0.11% % 0.10% R n GDP n % 0.11%

29 Varying Tax Dispersion Worker welfare is maximized when tax dispersion is eliminated Doubling tax dispersion from the current scenario would reduce welfare by 0.4% keeping government spending constant Still below E.U. levels of dispersion Figure : Welfare E ect of Varying Tax Dispersion Percentage Change in Worker Welfare Percentage Change in Standard Deviation in Tax Rates Across States Constant Spending Variable Spending

30 Alternative Fundamentals Welfare gains decrease as we increase the cross-state correlation between initial keep-tax rates and the total number of workers Whether a harmonized tax system that keeps government spending constant is superior to an observed spatial tax distribution depends on the specific country in question. Case Welfare GDP RankCorr(1 T n, L n)= % -0.14% Benchmark 0.15% 0.12% RankCorr(1 T n, L n)=1-1.07% 0.64%

31 Conclusion We study the e ects of the state tax distribution on welfare and the distribution of economic activity in the U.S. Quantitative economic geography model with U.S. state tax system Estimate key parameters using observed variation in taxes Simulate counterfactual state tax distributions keeping government spending constant n the U.S., tax dispersion leads to aggregate worker welfare losses across many specifications 0.2% worker welfare and 0.1% GDP gain from eliminating spatial dispersion in taxes accounting for 4% of GDP F Gains increase to 0.45% under progressive income taxes Doubling tax dispersion leads to welfare loss of 0.4% F F eliminating tax dispersion maximizes worker welfare the answers depend on the country in question

32 Distribution of Tax Rates Across States Density State Tax Rates in 2010 Sales Corporate ndividual ncome Sales Apportioned Corporate back

33 Tax Revenue as Share of GDP Across States percentiles of the distribution of sales, personal income, and corporate tax revenue shares: 76%-30%, 49%-0%, 8%-0% AK DE NH WY TX SD CO LA NV OR GA TN VA OK MT L MO WA ND FL AZ NE UT SC AL OH A NM MD NC PA KS N D NJ MA R CA KY M CT NY W VT MN AR MS ME WV H State Tax Revenue as Share of GDP in ncome Corporate Sales back

34 Dispersion in Sales Tax Rates Across States Sales tax rates: 5HighestRates Rate 5LowestRates Rate California 8.25% Delaware 0% New Jersey 7% Montana 0% Mississippi 7% New Hampshire 0% Tennessee 7% Alaska 0% ndiana 7% Oregon 0% Only final consumption is subject to sales taxes Firms do not pay sales taxes when buying intermediates back

35 Dispersion in ndividual ncome Tax Rates Across States ndividual income tax rates: 5 Highest Rates Rate 5 Lowest Rates Rate Oregon 6% Washington 0% Hawaii 5.2% Wyoming 0% North Carolina 5% Florida 0% Wisconsin 4.7% Nevada 0% Kentucky 4.7% Texas 0% back Average rate for median family type from NBER TAXsim dataset State income rates are applied after federal income tax Average federal income rate for median family is 9.2%; federal payroll tax rate is 7.2%.

36 Dispersion in Corporate Tax Rates Across States back Corporate tax rates: 5HighestRates Rate 5LowestRates Rate owa 12% Washington 0% Pennsylvania 10% Wyoming 0% Minnesota 10% Florida 0% New Jersey 9% Nevada 0% Rhode sland 9% Texas 0% Applied on profits and apportioned through sales, payroll, and capital Asingle-plantfirmj from i with export share s j ni to state n pays: F ti l j i to state i, whereti l = t corp i i l F s j ni tx n j i to state n, tn x = tn corp n x F t j j i in total, where t j = t corp fed + ti l + P n sj ni tx n nternal trade matters for determining the corporate tax rate. Federal corporate tax rate is 18%.

37 Estimation of {" F, F } Firm share in state n at time t: ln M nt = " F / ( 1) 1+ F F ( 1) ln ((1 tn) MPnt )+ " F F Rnt 1+ F F ( 1) ln P nt " F 1+ F F ( 1) ln cnt + M t + M n + M nt Observed covariates: c nt = MP nt = X n 0 w 1 nt n E n 0 t r n nt n P 1 nt n 0 nt P n 0 t t n 0 nt 1 n 1 Unobserved: M t 1 / ln t M n + M nt / ln (z nt)

38 Estimation of {" F, F } Firm share in state n at time t: ln M nt = " F / ( 1) 1+ F F ( 1) ln ((1 tn) MPnt )+ " F F Rnt 1+ F F ( 1) ln P nt " F 1+ F F ( 1) ln cnt + M t + M n + M nt dentification assumption: where E[ M nt (Z L nt, MP nt, M, n 0 nt M ) 0 ]=0 1, MP nt = X n 0 6=n E n 0 t P n 0 t t n 0 nt 1 back

39 Calibration of Technologies and Ownership Rates Production technologies { n, n o P n} match input and labor shares n n X n, wnln nx n Parameter Notation Match in the Data Average X ntermediates share in sales 1 n = n VA n 1 X n 0.62 w Labor share in value added 1 n = nl n 1 nx n 0.68 Asset positions {b n} are set to match trade deficits: apple n b n = +R + t corp ( tn x PnQ n ) ( 1) (1 n X fed n n) tn l back

40 Calibration of Fundamentals nverting the import-share equation: A in = in w ( 1)apple 1 1 n L ( 1)apple 2n 1 n w i L apple 3 i 1 where A in / H n n n zn A in A u A n ui A v F n (1 n)+ F! 1 This rationalizes in as equilibrium outcome mplies that s in is also matched given how { n, n, b n} are calibrated mplies that {w n, L n} can be rationalized as an equilibrium outcome because the market-clearing conditions that determine them are satisfied: X in = 1 for all i, n X s in = 1 for all n. i Note that breakdown of A in into each fundamental is not identified from here. back

41 Establishment and GDP Shares: Model vs. Data Model State GDP Share L PA OH NJ GA MA M VA NC WA CO AZ N KY LA CT MO W TN MN MD AL A AR KS ME DE MS MT NH H NM ND NENV OK SC OR UT VT WY SD R WV FL NY TX CA Model Establishment Share L OH PA M NJ GA NC N KYLA TN MA WA VA CT AL MD AZ W MN MO CO AR KS AOKOR SC MS DE HD MT NH ME NM ND SD R NENV UT WY WV VT TX NY FL CA Actual State GDP Share in 2007 Note: Slope is 1 (0). R-squared is 1. (a) State GDP Share Actual Establishment Share in 2007 Note: Slope is 1.01 (.04). R-squared is.92. (b) Share of Establishments back

42 Government Revenue Shares: Model vs. Data Model Tax Revenue as Share of GDP NH D WV A MS ME AR MD KY MN CA VA PA VT R KS NM MA N AZ CT NJ ND LA MO OK NE SC NC NY GA OH UT W AL L MT OR CO FL M DE TN NV WA TX WYSD H Model Sales Tax Revenue Share DE MT NH OR M L MS AZ VT MDCA R ND CT NJ SC VA NM MN PA D N AR OK UT NE MA WV MO KY KS NY CO W GA AL OH ME H NC A LA TN SD FL NV TX WA WY Actual Tax Revenue from Modeled Taxes as Share of GDP in 2007 Note: Slope is 1.65 (.14). R-squared is.74. (c) Tax Revenue Share of GDP Actual Sales Tax Revenue Share in 2007 Note: Slope is.83 (.05). R-squared is.85. (d) Sales Tax Share of Revenue Model ncome Tax Revenue Share NH TN FL NV SD TX WA WY MT H DE CO NY UT NC ME GA VA D SC MO AR MA WV OK NE W CT KY ALKS MD MSNM M A LA OH R CA L MN AZ N PA VT ND NJ OR Model Corporate Tax Revenue Share DE A LA TN OH N AL MNPA MT WKS NJKY ND MONE ORVT MA WV GA AR MEAZ CA CO CT NC FLOKNM D R L H SC MD NY VAMS UT M NV TX WA WY SD NH Actual ncome Tax Revenue Share in 2007 Note: Slope is.64 (.03). R-squared is.88. (e) ncome Tax Share of Revenue Actual Corporate Tax Revenue Share in 2007 Note: Slope is 1.29 (.21). R-squared is.45. (f) Corp. Tax Share of Revenue back

43 Price Distortion Sales apportionment leads to price distortion: where c i p ni (z) = ni t ni 1 z, where t ni = tx n P n 0 t x n 0s n 0 i 1 t i t ni =0undernodispersioninsalesapportionmentacrossstates(tn x = t x for all n) back

44 Dispersion in Worker and Firm Keep Rates Firm Keep rate: 1 (t corp fed + t l i + P N n=1 tx n s ni ) Worker Keep rate: 1 T n = 1 tt n t y fed 1+t c n t w fed Figure : Dispersion in Worker and Firm Keep Rates Firm keep rate in Worker keep rate in 2007 back