Profit shifting in Brazil and the impact of tax havens

Rathke, Alex A. T.

doi:10.1590/1808-057x201910040

Acessibilidade / Reportar erro

Brasil

Revista Contabilidade & Finanças

Español English

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

Original Article • Rev. contab. finanç. 32 (85) • Jan-Apr 2021 • https://doi.org/10.1590/1808-057x201910040 copy

Profit shifting in Brazil and the impact of tax havens

Authorship SCIMAGO INSTITUTIONS RANKINGS

ABSTRACT

We investigate tax-induced profit shifting in Brazil and the impact of tax havens on the shifting behavior of firms. Profit shifting research in Brazil is virtually non-existent, although the shifting incentives in Brazil are prominent. Our research fills this gap with evidences in the novel Brazilian context. Profit shifting is a tax-minimization strategy where multinational enterprises perform intra-firm transactions to allocate taxable profits to low-tax locations. Brazil combines a remarking set of profit shifting incentives, especially a high corporate tax rate, extremely complex tax system, and distinguished transfer pricing rules. Further researches may leverage from the shifting incentives in Brazil, since it provides opportunities to investigate additional factors that affect the shifting behavior of firms. We analyze 989 transaction-by-country observations for the period of 2010-2017. Baseline analysis follows the robust least squares approach with controlling covariates. Linear estimate model derives from the conventional Cobb-Douglas production function, to analyze the impact of shifting incentives on profit maximization. We find that Brazilian firms have a high level of intra-firm transactions with related parties located in low-tax countries, especially with tax havens. It represents a strong evidence of profit shifting behavior in Brazilian firms.

Keywords:
profit shifting; transfer pricing; tax havens; base erosion and profit shifting (BEPS)

Variables	Obs. (n)	Mean	SD	Min.	Max.
Firm-level
Volume of intra-firm transactions	989	468.4	4,030.7	0.0	57,026.0
Labor expenses	648	1,180.23	2,263.61	2.1	29,732.1
PPE	989	4,926.0	32,768.3	0.0	629,830.9
Inventories - Net	987	1,864.2	3,118.3	3.6	29,057.2
Revenues - Net	989	17,479.0	36,270.2	245.7	321,638.0
Assets - Total	989	19,163.5	52,835.9	145.8	900,135.1
Capital expenditures	984	1,033.3	4,198.3	3.0	71,311.0
Liabilities - Total	989	12,593.6	35,738.6	74.1	645,403.9
Costs and expenses - Total	989	15,801.0	33,250.5	226.8	259,224.0
Country-level
Income tax rate- Foreign (%)	336	27.0	8.1	0.0	35.0
Tax differential (%)	336	7.0	8.1	-1.0	34.0
GDP (US$ 000.000.000)	336	3,903.2	5,963.3	3.4	19,390.6
GDP ratio (US$)	336	1.8	2.8	0.0	10.4
Distance from Brazil - Regularized	42	0.4072	0.2411	0.0672	1.0000
Foreign countries (n)	42
Tax havens (n)	18
Years (n)	8
Industries (n)	7

	Estimate coefficients: γ, β_l
Variables	1		2		3		4
Tax Param.: log(f(Δτ)) = Δτ	0.1330	*	0.1404	*	0.1683	*	0.1917	*
Tax Param.: log(f(Δτ)) = Δτ	0.0404		0.0411		0.0466		0.0292
Labor expenses	-0.0394		-0.0230		-0.0418		-0.0285
Labor expenses	0.0228		0.0252		0.0234		0.0249
PPE	0.0723	*	0.0081		0.0720	*	0.0065
PPE	0.0228		0.0283		0.0231		0.0191
Inventories	-0.1061		-0.0142		-0.1070		-0.0102
Inventories	0.0599		0.0495		0.0594		0.0457
Industry-level controls	N		Y		N		Y
Country-level controls	N		N		Y		Y
Observations (n)	638		638		638		638
F-statistics	16.30	**	14.33	**	12.27	**	13.37	**
R-squared adjusted	0.0877		0.1730		0.0959		0.1889

	Estimate coefficient: γ’
Tax Param.: log(f(Δτ))	1		2		3		4
1. log(1 + Δτ)	0.1428	*	0.1497	*	0.1845	*	0.2096	*
	0.0280		0.0444		0.0510		0.0327
	(13.37)	**	(14.09)	**	(12.01)	**	(13.18)	**
2. Δτ, direct linear reg.	0.1571	*	0.1666	*	0.1962	*	0.2245	*
	0.0490		0.0502		0.0565		0.0344
	(16.78)	**	(14.48)	**	(12.37)	**	(13.36)	**
3. 1/(τ + 0.1)	0.0105	*	0.0112	*	0.0114	*	0.0131	*
	0.0033		0.0034		0.0035		0.0037
	(17.02)	**	(15.02)	**	(12.15)	**	(13.28)	**
Firm-level covariates	Y		Y		Y		Y
Industry-level controls	N		Y		N		Y
Country-level controls	N		N		Y		Y
Observations (n)	638		638		638		638

	Estimate coefficients: γ, μ
Variables	1		2		3		4
Tax Param.: log(f(Δτ))	0.1769	*	0.1938	*	0.0116	*	0.2070	*
Tax Param.: log(f(Δτ))	0.0473		0.0518		0.0035		0.0578
Tax Haven Param.: T∈{0,1}	0.0214	*	0.0216	*	0.0201	*	0.0255	*
Tax Haven Param.: T∈{0,1}	0.0053		0.0053		0.0052		0.0063
Firm-level covariates	Y		Y		Y		Y
Industry-level controls	Y		Y		Y		Y
Country-level controls	Y		Y		Y		Y
Observations (n)	638		638		638		638
Obs. tax havens (n)	392		392		392		392
F-statistics	14.52	**	14.39	**	14.14	**	0.1457	**
R-squared adjusted	0.2163		0.2146		0.2114		0.2169

	Estimate coefficients: γ, μ, μ’
Variables	1		2		3		4
Tax Param.: log(f(Δτ))	0.1300	*	0.1408	*	0.0079	**	0.1466	*
Tax Param.: log(f(Δτ))	0.0331		0.0362		0.0033		0.0377
A.Tax Haven Par.: T∈{0,1}	0.0457	*	0.0474	**	0.0427		0.0567	**
A.Tax Haven Par.: T∈{0,1}	0.0204		0.0203		0.0245		0.0250
C.Tax Haven Par.: T’∈{0,1}	0.0175	*	0.0174	*	0.0173	*	0.0204	*
C.Tax Haven Par.: T’∈{0,1}	0.0052		0.0052		0.0052		0.0061
Firm-level covariates	Y		Y		Y		Y
Industry-level controls	Y		Y		Y		Y
Country-level controls	Y		Y		Y		Y
Observations (n)	638		638		638		638
Obs. A.Tax Havens (n)	57		57		57		57
Obs. C.Tax Havens (n)	335		335		335		335
Fstatistics	14.04	***	14.01	***	13.44	***	14.20	***
R-squared adjusted	0.2227		0.2224		0.2146		0.2249

	Estimate coefficients: γ, μ, μ’
Variables	1		2		3
Tax Param.: log(f(Δτ)) = Δτ	0.1000	*	0.0944	*	0.1049	*
Tax Param.: log(f(Δτ)) = Δτ	0.0202		0.0205		0.0218
A.Tax Haven Par.: T∈{0,1}	0.0282		0.0320	**	0.0288	**
A.Tax Haven Par.: T∈{0,1}	0.0144		0.0139		0.0132
C.Tax Haven Par.: T’∈{0,1}	0.0111	*	0.0114	*	0.0111	*
C.Tax Haven Par.: T’∈{0,1}	0.0034		0.0035		0.0035
Firm-level covariates - Type	CAPEX		FS		CS
Industry-level controls	Y		Y		Y
Country-level controls	Y		Y		Y
Observations (n)	974		979		979
Obs. A.Tax Havens (n)	53		57		57
Obs. C.Tax Havens (n)	335		335		335
F-statistics	18.49	***	15.73	***	17.21	***
R-squared adjusted	0.1774		0.1742		0.1773
Wald test - Fstatistics	8.19	***	6.28	***	6.25	***

Note: This table presents the estimate coefficients γ, μ,μ’ for the regression models of different firm-level covariates in matrix X, excluding outlier observations from the sample. Numbers in plain refer to the estimate coefficients. Numbers in italic refer to the estimates’ standard errors. Standard errors are obtained from the White's covariance matrices to account for heteroscedastic consistency. Outliers are selected based on the traditional Bonferroni adjustment on studentized residuals of preliminary baseline estimates. Derivation of the baseline model is presented in section 3. Details about all variables are presented in section 3 and Table 1. Column 1 presents the estimate coefficients including the amount of capital expenditures (CAPEX) of each firm, for each year. Column 2 presents the estimates including the components of the financial structure (FS) of each firm, for each year, and is composed by the amounts of the short-term leverage, long-term leverage, and shareholders’ equity. Column 3 presents the estimates including the components of the operational cost structure (CS) of each firm, for each year, and is composed by the amounts of operational costs and operational expenses. A.Tax Haven parameter T refers to an indicator variable with value 1 if the foreign related country is classified as an absolute tax haven and 0 otherwise. C.Tax Haven parameter T’ refers to an indicator variable with value 1 if the foreign related country is classified as a conditional tax haven and 0 otherwise. Countries are classified as absolute or conditional tax havens according to the Brazilian tax rule Instrução Normativa da Receita Federal do Brasil (IN RFB) n. 1037/2010. Estimates are computed with respect to the monotone transformation log(1 + variable) for the dependent variable and the firm-level covariates. Fixed controls for industry effects refer to an indicator variable for the firm’s industry sector. Controls for country effects refer to the ratio between annual GDP of the foreign country and the annual GDP in Brazil and the distance between Brazil and the foreign country. Wald test assesses the existence of significative covariates.
Y = yes.
* tstatistic significative at < 0.01 level; ** tstatistic significative at < 0.05 level; *** F-statistic significative at < 0.01 level.
Source: Elaborated by the author.

Universidade de São Paulo, Faculdade de Economia, Administração e Contabilidade, Departamento de Contabilidade e Atuária Av. Prof. Luciano Gualberto, 908 - prédio 3 - sala 118, 05508 - 010 São Paulo - SP - Brasil, Tel.: (55 11) 2648-6320, Tel.: (55 11) 2648-6321, Fax: (55 11) 3813-0120 - São Paulo - SP - Brazil
E-mail: recont@usp.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] Email: alex.rathke@alumni.usp.br