Tax Policy, Firm Size and Effective Tax Rates: Empirical Evidence from Quantile Regression. Yao-Chih Hsieh *

Transcription

1 INTERNATIONAL JOURNAL OF ECONOMICS Volume 5 Number 2 December 2011 pp Tax Policy, Firm Size and Effective Tax Rates: Empirical Evidence from Quantile Regression Yao-Chih Hsieh * In order to promote economic development in Taiwan, Taiwan government used a variety of tax policies to stimulate investment through tax reductions. This article examines the relationship between firm size and effective tax rates under Promotion Industrial Upgrading Act (PIUA). Both OLS regression and quantile regression technique are employed to analysis this. The findings from the empirical results are as follows. The PIUA policy supports political power hypothesis.. Firms with moderate high ETRs get the largest tax reductions from PIUA when firm size becomes larger. JEL Classification: H25, H32. Keywords: Effective tax rate; Quantile regression; Political power (cost) hypothesis. 1. INTRODUCTION By implementation of the Promotion Investment Act (PIA) before 1992 and the Promotion Industrial Upgrading Act (PIUA), also known as the Statute for Upgrading Industries) from 1992 to 2009, Taiwan government offered a variety of tax preferences to industries and firms. PIA focused on fostering the emerging industries and encouraging expansion of export. PIUA provided tax exemptions and incentives to encourage companies engaged in research and development, personnel training, the establishment of international brand, purchase of automation equipment, and pollution control. The so-called miracle of Taiwan s economic development and the well-established high-tech industry resulted from these tax policies. Under these two acts, big business conglomerates have saved more than NT $1.5 trillion in tax payment during the past 50 years. After the expiration of the Statute for Upgrading Industries on Dec. 31, 2009, a so-called Industrial Innovation Bill was proposed by Taiwan s Executive Yuan (Cabinet) to continue attracting investment capital from Taiwan-based firms and those from abroad to create job opportunities and generate more tax revenues. The Bill says that enterprises with investment for industrial innovation, and those cultivating talent to create jobs may be eligible to receive tax incentives. At last, the business income tax rate is cut to 17 % from 25 %. Originally, clauses in the bill stipulate that only those among the world s 500 biggest enterprises with operations in Taiwan can apply for the 15 per cent corporation income tax break. This is the incentive to foster international competitiveness such as for promoting R&D or establishing operational headquarters in Taiwan. Since it was widely conceived that the bill was customized to favor foreign conglomerates, the plan has provoked strong suspicion and complaints from some business and industry leaders. At the last stage, the Taiwan Cabinet * Department of Public Finance, Feng Chia University, Taichung 40724, Taiwan, ROC, (ychsieh@fcu.edu.tw)

2 228 Yao-Chih Hsieh decided that all local small and medium-sized businesses irrespective of their size or the sectors in which they operate can apply for the incentives suggested in the bill. Among the intense debates, two crucial and opposing extreme arguments are why favor foreign enterprises and why favor large-scale firms. A common saying is that the large-scale firms in Taiwan have enjoyed the tax incentives more than the small and medium-sized businesses. This is an issue concerning fairness and tax neutrality. However, previous study (Kim 1998) pointed out that lager firms pay higher effective tax rates (ETRs) before 1992 in Taiwan. The results were opposed to what we mentioned earlier. Success in the political process normally leads to future changes in effective tax rates. Will the larger firms still get higher effective tax rates under PIUA? In addition, not all firms with the same size or with different firm size have equal access to these tax codes; therefore, firms with the same size or with different firm size could have different marginal effects on effective tax rates. This paper is to provide empirical evidences to answer those two questions. Variability of ETRs among firms has been recognized in literature. Originated from a serial reports published by the Citizens for Tax Justice (CTJ 1984, 1985, 1986) that the largest United States corporations were not paying their fair share of taxes. Whether the ETRs are systematically related to firm size has been the focus of the tax policy debates and literature studies. Theoretically, there are two different hypotheses suggested in literature to explain the relation between firm size and level of ETRs. One, the political cost hypothesis, is that the higher visibility of larger and more successful firms might be victims of greater regulator actions and wealth transfers and as a result, incur a political cost in the form of a higher ETR (Zimmerman 1983). Another is that larger firms pay less tax because they devote more efforts to tax planning and political lobbying so that the large firms pay their share of tax less. This is referred to as the political power (or clout) hypothesis. Empirically, studies on the relation between ETRs and firm size have produced conflicting results. Zimmerman (1983) observes a positive association between ETRs and firm size while Porcano (1986) observes a negative association. No association between ETRs and firm size is found in Stickney and McGee (1982) and Shevlin and Porter (1992). Subsequent studies have tried to reconcile the conflicting results by using modified proxies, time period, data basis, and methodologies (e.g., Gupta and Newberry, 1997; Kern and Morris, 1992; Wilkie and Limberg, 1990; Holland, 1998; Kim and Limpaphayom, 1998; Derashid and Zhang, 2003; Liu and Cao, 2007). All findings from the existing literature results from using OLS in which the conditional mean of ETRs on firm size are studied. Only mean relation between ETRs and firm size is explored. Those conclusions just told us that large firms faced higher or lower ETRs. It is one to one relationship. In fact, because not all firms have equal access to tax codes, firms with the same size could get different tax preferences; thus previous studies only discussed part of story about the relationship between firm size and effective tax rate. The characteristic of the firms may be heterogeneous in ways that are associated with size but not captured by effective tax rates. Since each firm size could have heterogeneous in the potential tax policy effects, we use a quantile regression approach. Quantile regression estimates are more robust than OLS estimates in the case because they place less weight on outliers in the firm size distribution. Besides, quantile regression is not only concerned with the ETR effect on the average individual firm size, but allow one to estimate the marginal effect of a given firm size on ETRs at different point in the conditional achievement distribution.. It is important to analyze the behavior of firm size in each quantile of the conditional effective tax rates when we discuss tax policies.

3 Tax Policy, Firm Size and Effective Tax Rates: Empirical Evidence from 229 The rest of this article is organized as follows. A brief introduction of quantile regressions is provided Section II. Section III presents the empirical data and empirical results. Conclusions are in the Section IV. 2. THE ECONOMETRIC FRAMEWORK Quantile regression is outlined as follows: y i = x i β τ + u τi and (1) Quantile τ (y i x i ) = x i β τ, (2) Where y i equals the dependent variable (effective tax rate) of firm i, x i equals a vector of independent variables (leverage, capital intensity, inventory intensity and return on assets) of firm i, β τ equals the vector of parameters associated with the τ th quantile (percentile), and u τi equals an unknown error term. Unlike ordinary least squares (OLS), the distribution of the error term u τi remains unspecified in Equation 2. We require only the conditional τ th quantile of the error term equals zero, that is, Quantile τ (u τi x i ) = 0 Quantile τ (y i x i ) = x i β τ equals the τth conditional quantile of y given x with τ (0,1). By estimating β τ using different values of τ, quantile regression permits different parameters across different quantiles of the dependent variable. In other words, repeating the estimation for different values of τ between 0 and 1, we trace the distribution of y conditional on x and generate a much more complete picture of how explanatory variables affect the dependent variable. Furthermore, instead of minimizing the sum of squared residuals to obtain the OLS (mean) estimate of β, the τ th quantile regression estimate β τ solves the following minimization problem: min β 2τ yi x iβ + 2(1) τ yi x i. β i (3) { : } { : } i i yi xiβ i i yi< x β Any quantile of the distribution of dependent variable conditional on independent variables can thus be obtained by changing continuously from zero to one and using linear programming methods to minimize the sum of weighted absolute deviations. Quantile regression estimates of Equation 1, so the coefficient estimates differ by quantile, whereas OLS estimates provide information on their values at the conditional mean of the dependent variable. 3. EMPIRICAL DATA AND ESTIMATED RESULTS The sample data used is collected from the Taiwan Economic Journal (TEJ) data base. It consists of 718 firms each year listed on Taiwan stock market in the nine-year period 1998 to There are several ways to measure ETR, but no acceptable one has yet emerged. The key issues include how to determine tax expenses, and how to calculate taxable income and the variability of ETR. Procano (1986) pointed out that his measure is superior because it better reflects a firm s ability to meet its obligations. This article follows the definition of Procano because we discuss the effects of tax policies. The ETR is defined as tax expenses (deferred tax provisions are not included) divided by profit before interest and tax paid. It is worthy to mention that the measurement issue regarding concerns firms with negative income or tax refunds; this article

4 230 Yao-Chih Hsieh follows Gupta and Newberry (1997) to constrain the ETR of our sample firms to lie between 0 and 1. To explore the marginal effect of firm size on ETRs, the following firm-specific characteristics are taken as control variables: leverage (total liabilities divided by total asset value, denoted as LEV ), capital intensity (net fixed assets divided by total assets, denoted as CI ), inventory intensity (inventory divided by total assets, denoted as II ), and return on assets (pre-tax profits divided by total assets, denoted as ROA ). The reasons to choose these variables are based on previous studies (e.g., Porcano (1986), Gupta and Newberry (1997), Derashid and Zhang (2003), Liu and Cao (2007)). The main explanatory variable, firm size (denoted as Size ) is measured as the natural logarithm of total asset value. The sample averages for all variables of each year used in this paper are presented in Table 1. The second column denoted as Zeros is the ratio of firms with zero ETRs among all 718 firms. It is worthy to note that there are more than 20 % of firms with zero ETRs each year from 1998 to This fact suggests that the considered quantiles are larger than 0.3. Table 1 Summary Statistics for All Variables Zeros ETR Size LEV CI II ROA The quantiles considered in this paper are 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, and 0.9. In addition to the estimations for conditional quantiles, the conditional mean is also estimated by OLS (ordinary least squares) estimator. The empirical estimated results are presented in Table 2 to Table 6. Table 2 Estimated Quantile Regressions Results for Size on ETR Size lm rq0.3 rq0.4 rq0.5 rq0.6 rq0.7 rq0.8 rq * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

5 Tax Policy, Firm Size and Effective Tax Rates: Empirical Evidence from 231 The OLS and quantile regression estimated results for the effects of firm size on ETRs are presented in Table 2. For the OLS estimated results shown in the second column labeled with lm, the estimated coefficients are all significant at 5 % type I error for each year from 1998 to The conditional quantile regression estimated results are shown in the columns labeled with rq0.3, rq0.4, rq0.5, rq0.6, rq0.7, rq0.8, and rq0.9 for quantiles 0.3, 0.4,0.5, 0.6, 0.7, 0.8, and 0.9, respectively. All estimated coefficients are statistically significant at 5 % Table 3 Estimated Quantile Regressions Results for LEV on ETR LEV lm rq0.3 rq0.4 rq0.5 rq0.6 rq0.7 rq0.8 rq * * * 0.106* * 0.087* * 0.055* 0.074* 0.058* * 0.060* 0.083* * Table 4 Estimated Quantile Regressions Results for CI on ETR CI lm rq0.3 rq0.4 rq0.5 rq0.6 rq0.7 rq0.8 rq * * * * * * * * * * * * * * * * * Table 5 Estimated Quantile Regressions Results for II on ETR II lm rq0.3 rq0.4 rq0.5 rq0.6 rq0.7 rq0.8 rq * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

6 232 Yao-Chih Hsieh Table 6 Estimated Quantile Regressions Results for ROA on ETR ROA lm rq0.3 rq0.4 rq0.5 rq0.6 rq0.7 rq0.8 rq * 0.288* 0.325* 0.306* 0.251* 0.133* 0.100* 0.121* * 0.317* 0.354* 0.346* 0.322* 0.187* 0.047* 0.123* * 0.319* 0.354* 0.372* 0.341* 0.296* * 0.272* 0.345* 0.393* 0.336* 0.278* 0.174* * 0.297* 0.408* 0.476* 0.429* 0.340* 0.249* * 0.324* 0.478* 0.527* 0.470* 0.404* 0.153* * 0.213* 0.359* 0.408* 0.411* 0.377* 0.205* 0.094* * 0.216* 0.374* 0.429* 0.383* 0.301* 0.256* * 0.317* 0.391* 0.377* 0.395* 0.379* 0.208* 0.139* type I error and support the political power hypothesis. That is the marginal effect of firm size on ETRs is significantly negative no matter in the conditional mean or in the conditional quantiles for each year from 1998 to Both the results of OLS and quantile regressions support the political power hypothesis. The results also indicate that the larger firms in Taiwan get the success in the political process. In addition, Table 2 indicates that there strong evidence that the effect of firm size is not constant, but varies among the various quantiles. All the estimates have a negative sign, meaning that when firm size increase the ETRs decrease. It is important to note that the absolute value of the coefficients is larger for quantile 0.6 or 0.7. This shows that firm size is relatively efficient in decreasing ETR at the moderate high of the conditional ETRs. We can infer that larger firms with moderate high ETRs (quantile 0.6 or 0.7) get the largest tax reductions from PIUA and organize their activities to achieve optimal tax savings when firm size becomes larger. 4. CONCLUSIONS The purposes of both Promotion Investment Act (PIA) and Promotion Industrial Upgrading Act (PIUA) are to reduce the tax burden of firms. PIA focuses on emerging industries and encouraging expansion of export; PIUA focuses on encouragement companies engaged in research and development, personnel training, the establishment of international brand, purchase of automation equipment, and pollution control. Previous study pointed out that PIA supported political cost hypothesis. The results of this article support political power hypothesis under PIUA. In this paper, the regression model for ETRs with firm characteristics, i.e., firm size, leverage level, capital intensity, inventory intensity, and return on asset, is estimated using OLS and quantile regression estimators. The sample observations considered in this paper are the 718 firms listed in the Taiwan stock market from 1998 to Three main findings are obtained: (1) the political power hypothesis is supported by the empirical results that the marginal effects of firm size on ETRs from OLS and quantile regressions are significantly negative at 5 % type I error. Both results support political power hypothesis under PIUA. (2) The estimated marginal effects of quantile regressions indicate firm size on ETRs are relative small at both lower and higher quantiles and relative large at middle quantiles. (3) Firms with moderate high ETR will get the largest tax reductions from PIUA when firm size is getting larger.

7 Tax Policy, Firm Size and Effective Tax Rates: Empirical Evidence from 233 References Buijink, W., B. Janssen and Y. Schols (2002), Evidence of the Effect of Domicile on Corporate Effective Tax Rates in the European Union, Journal of International Accounting, Auditing, and Taxation, 11, Bofinger, E. (1975), Estimation of A Density Function Using Order Statistics, AustralianJournal of Statistics, 17, 1-7. Callihan, D. S. (1994), Corporate Effective Tax Rates: A Synthesis of the Literature, Journal of Accounting Literature, 13, Derashid, Check, and Hao Zhang (2003), Effective Tax Rates and the Industrial Policy Hypothesis: Evidence from Malaysia, Journal of International Accounting Auditing and Taxation, 12, Fitzenberger, B. (1998), The Moving Blocks bootstrap and Robust Inference for Linear Least Squares and Quantile Regressions, Journal of Econometrics, 82, Gupta, S., and K. Newberry (1997), Determinants of the Variability of Corporate Effective Tax Rates: Evidence from Longitudinal Data, Journal of Accounting and Public Policy, 16, Holland, K. (1998), Accounting Policy Choice: The Relationship between Corporate Tax Burdens and Company Size, Journal of Business Finance and Accounting, 25, Huang, G., and M. Song (2004), The Determinants of Capital Structure: Evidence from China, Quarterly Economic Journal, 2, Kim, K. A. and P. Limpaphayom (1998), Tax and Firm Size in Pacific-Basin Emerging Economies, Journal of International Accounting Auditing and Taxation, 7, Kern, B. B. and M. H. Morris (1992), Taxes and Firm Size: The Effect of Tax Legislation during the 1980s, Journal of the American Taxation Association, 14, Koenker (2005), Quantile regression, Cambridge University Press. Koenker, R. and B. J. Park (1996), An Interior Point Algorithm for Nonlinear Quantile Regression, Journal of Econometrics, 71, Liu, Xing and Shujun Cao (2007), Determinants of Corporate Effective Tax Rates Evidence from Listed Companies in China, The Chinese Economy, 40, MacKie-Mason, Jeffrey K. (1990), Do Taxes Affect Corporate Financing Decisions? Journal of Finance, 45, Manzon, G., and W. R. Smith (1994), The Effect of the Economic Recovery Tax Act of 1981and the Tax Reform Act of 1986 on the Distribution of Effective Tax Rates, Journal of Accounting and Public Policy, 13, Plesko, George (2003), An Evaluation of Alternative Measures of Corporate Tax Accounting and Economics, 35, Rate, Journal of Porcano, T. (1986), Corporate Tax Rates: Progressive, Proportional, or Regressive, Journal of Accounting and Public Policy, 7, Shelvin, T., and S. Porter (1992), The Corporate Tax Comeback in 1987: Some Further Evidence, Journal of American Taxation Association, 14, Si, Donghui (2003), Equity and Competition in Transform Research: Evidence from Listed Companies in China, Economic Research, 8, Siegfried, J. (1974), Effective Average U.S. Corporate Income Tax Rates, National Tax Journal, 27,

8 234 Yao-Chih Hsieh Stickney, C., and V. McGee (1982), Effective Corporate Tax Rates: The Effect of Size, Capital Intensity, Leverage, and Other Factors, Journal of Accounting and Public Policy, 1, Wang, Fang (1999), A Study on the Income Tax Burden of Listed Companies in Research, 5, China, Economic Wilkie, P. (1988), Corporate Average Effective Tax Rates and Inferences About therelative Tax Preference, Journal of the American Taxation Association, 10, Wilkie, P. J. and S.T. Limberg (1990), The Relationship Between Firm Size and Effective Tax Rate: A Reconciliation of Zimmerman (1983) and Porcano (1986),Journal of American Taxation Association, 11, Yang, Zhigang, Ding Ling, and Bingzhen Wu (2000), Positive Research on Value Added Tax and Income Tax, Economic Research, 12, Zimmerman, J. (1983), Tax and Firm Size, Journal of Accounting and Economics, 5,