Cross-Sector Tax Discrimination and Economic Performance:

A Cross-Country Analysis

Young Lee*

Hanyang University

Taeyoon Sung**

Yonsei University

Taejong Kim***

KDI School of Public Policy and Management

June 2008

Abstract

Using corporate financial statement data in 70 countries, this paper measures tax burden in each sector and analyzes the impact of tax burden on the growth of sector and overall economy. We find that sectors and countries with lower effective tax rates tend to grow faster. In addition to the level of tax rates we also examine the effect of cross-sector differences in tax burden within a country, which can be seen as a proxy for industrial policy through taxation. We find that industrial policy through taxation hampers the growth of other sectors and fails to promote the economic growth, though it can promote the growth of the favored sector. We also find that negative effect of industrial policy is stronger in OECD countries, implying that the room for government intervention becomes smaller as the economy and the market develops.

Key words: Corporate Income Taxation, Industrial Performance, Economic Growth, Preferential Corporate Taxation

JEL classification: H21; E62

______

* Corresponding Author. Professor Young Lee, College of Economics and Finance, Hanyang University, Seoul 133-791, Korea. E-mail:

** Professor Taeyoon Sung, School of Economics, Yonsei University, Seoul, 130-722, Korea. E-mail:

*** Professor Taejong Kim, KDI School of Public Policy and Management, Seoul, 130-868 Korea. E-mail:

I. Introduction

We attempt to measure cross-sector discrimination in corporate income taxation and analyze its impact on both industry-level performance and overall economic growth. For this purpose, we utilize a large-scale international panel data of firm-level financial statements. Given the data set, we can estimate effective marginal tax rates cross industries, and examine the relationship of the estimated rates with economic performance. In particular, since we have estimated effective marginal tax rates even at the industry level, we can examine the implications of cross-sector differences, as well as the absolute level of corporate tax burden, on economic performance.

The discriminatory application in corporate taxation, such as preferential tax rates, depreciation rules, and tax credits for investment and R&D expenditures, is a widely-adopted policy measure across countries. This measure can be interpreted to represent a kind of industrial policy through taxation. Theoretically, the overall impact of cross-sector tax discrimination on economic growth is ambiguous. First, the discrimination policy in taxation can distort resource allocation, and thus hampering growth.[1] On the other hand, the policy may compensate for positive externalities from tax-favored sectors to the rest of the economy. Thus, in this case, a prerequisite for the effectiveness of tax favors is the ability of the government to identify appropriate industries and set tax rates at optimal levels. Thus, figuring out whether the discriminative tax policy promotes economic growth calls for an empirical investigation.

The rest of the paper is organized as follows. In Section 2, we begin by reviewing the literature on the relationship between taxation and economic growth, and survey various concepts of the effective tax rate. The review also summarizes the literature on corporate tax burden and economic growth. Section 3 measures effective marginal tax rates across sectors for each country. To capture the effective marginal tax rate in each industry, we estimate coefficient from regressing corporate income taxes on before-tax earnings of firms in a given sector. For this purpose, we employ the OSIRIS dataset, a large-scale international longitudinal database of corporate financial statements.[2] Section 4 analyzes how estimated industry-level tax differences affect industry-level growth. Finally, Section 5 studies their impact on overall economic growth. The analysis in Section 5 incorporates the impact of statutory tax rates as well as the estimated effective tax rates. For the analysis in Section 5, we combine the estimated effective tax rates, the aggregate data from the World Bank's World Development Indicators, and statutory tax rates coming from the Worldwide Summary of Corporate Taxes from Price Waterhouse. Section 6 provides concluding remarks.

2. Literature Review

As already mentioned, distortion in taxation can disturb efficient resource allocation, and thus, hamper capital accumulation and growth in the overall economy (i.e., see Feldstein, 1974, 1978; Chamley, 1981; Becker, 1985; Judd, 1985; Jones and Manuelli, 1990; Rebelo, 1991). However, if taxation appropriately compensates for positive externalities from some sectors to the rest of the economy, taxation may contribute to efficient allocation of resources and economic growth.

Thus, the theoretical perspectives call for empirical investigation on the effect of taxation on growth. However, in literature, empirical studies report mixed results. For example, King and Rebelo (1990) and Jones, Manuelli, and Rossi (1993) report that corporate income taxation reduces economic growth. On the other hand, some studies, such as Lucas (1990) and Stokey and Rebelo (1995), suggest that the effects are either insignificant or may even work in the opposite direction. They are mostly calibration studies based on US data.

Ambiguity in calibration results also call for empirical studies that attempt to trace the impact of taxation in the growth regression framework by using the cross-country data. These studies include Skinner (1987), Koester and Kormendi (1989), Easterly and Rebelo (1993), Dowrick (1996), Agell, Lindh, and Ohlsson (1997), Bibbee, Leibfritz, and Thornton (1997), Mendoza, Milesi-Ferretti and Asea (1997). These studies suggest either negative or insignificant impact of overall corporate income taxation on economic growth. More recently, Lee and Gordon (2005) utilized a newly-available data on statutory corporate income taxes, and find a negative and significant effect of corporate income tax rates on economic growth. Not only were the results robust to the inclusion of other growth factors, but also their instrumental variable and fixed effects estimates confirmed the basic results.

In contrast to the earlier studies on the relationship tax and economic growth, this paper focuses on effective tax rates and cross-sector differences in corporate income taxation, and their impact on economic performance. Hence, this study supplements the existing literature in the sense that corporate taxation may affect overall growth not just through the aggregate tax rate overall, but also through the preferential treatment of certain industries. Cross-sector tax discrimination may disturb the equalization of before-tax rates of return across industries, and thus hamper the efficient resource allocation and capital accumulation.

3. Measuring Industry-Level Tax Burden

The main data analyzed in this paper come from the OSIRIS database provided by Bureau van Dijk Electronic Publishing. OSIRIS reports financial data for 47,180 business corporations from 1978 to 2006 from 138 countries. Due to missing values and entry & exit of firms, the actual number of firms in the data is smaller. For example, the number of firms with non-missing value of before-tax earnings is around 30,000 in 2005 from 107 countries. By using the financial data of business corporations, we construct industry-level and country-level variables. We generate four 5-year sub-samples: late 1980's (1986-2000), early 1990's (1991-1995), late 1990's (1996-2000), and early 2000's (2001-2005). For these four periods, we construct industry-level and country-level time-series, cross-section data. Our data, industry-level or country-level, is an an unbalanced panel.

Our key variable, effective tax rate (hereafter, ETR), comes from regressing corporate income taxes on before-tax corporate earnings. We followed method used in Koester and Kormendi (1989) which estimate effective marginal tax rates for a country by regressing government's tax revenue on GDP for a sample of countries. To prevent outliers from exercising undue influence on our estimates, we adopt median regressions for our purpose. Reported before-tax earnings are adjusted for possible loss-carry-forward and loss-carry-backward. When we estimate ETR, we use only firms with positive (loss-carry-adjusted) earnings and positive tax payments, which results in around 30% decrease in the number of firms. Using only firms with normal operation improves the fit considerably.

Our estimated ETR reflect not just statutory marginal tax rates (hereafter, STR) but also a myriad of provisions such as rules applying to investment and R&D tax credits and special tax exemptions.[3] Our estimation is an attempt to capture the overall impact of statutory tax rates, tax credits and special tax exemptions.

Our country-level estimates of ETR are strongly correlated with statutory corporate tax rates. The correlation coefficient between ETR and STR is 0.50 with p-value 0.000. <Figure 1> shows how ETR and STR are related for each country in the 2001-2005 sub-sample. If the estimated effective tax rates coincide with statutory tax rates, observations come along with the 45-degree. Although there are some cross-country differences, overall, the estimated ETRs turn out to be consistent with STRs.

Not all the observations appear exactly on the 45-degree line. Some countries appear above the 45-degree line, while other countries are located below the 45-degree line. For the countries above the 45-degree line, the estimated effective tax rates are higher than statutory tax rates. Given that our statutory tax rates are rates by the central government, we can observe this pattern if local government tax burden on corporate income is substantial. This pattern is observed in several OECD countries, such as Norway, Switzerland, Japan, Germany, Italy, and Ireland. We also observed that ETR is higher than STR in some countries with zero statutory rates, such as Jordan, Bermuda, Cayman Islands, Saudi Arabia, and Kuwait.

On the other hand, when various tax exemptions are provided, the estimated effective tax rates can be lower than statutory tax rates. In this case, observations can appear below the 45-degree line. We observe that ETR is lower than STR by more than 10%p in Egypt, Ecuador, Paraguay, Latvia, Belgium, Taiwan, Sri Lanka, Israel, Philippines, Luxembourg, Austria, and Pakistan.

We examine how effective tax rates changed over time for three groups of countries: all, OECD countries, and non-OECD countries. Since our data is an unbalanced panel, we cannot use a simple average of estimated ETR, and need to control for missing values. Using country-level data, we run simple regressions of estimated ETR on period dummies and country dummies. <Figure 3> shows that, on average, our estimated ETR decreased by 8%p from 32% in the late 1980s to 24% in the early 2000s.[4] Decrease in effective tax rates is more salient in non-OECD countries. The average ETR in non-OECD countries dropped by 11%p, while those in OECD countries dropped only by 5%p. STR also dropped considerably over the same period. One interesting pattern is that the difference in ETR between OECD and non-OECD countries becomes much larger than that in STR, suggesting that tax credits and tax exemptions become used more widely in non-OECD countries over time.

We estimate ETR not just for countries but also for industries in a given country. Our classification of industry is based on Standard Industry Code (SIC). Starting from one-digit SIC code, we combine or divide them to have similar numbers of observation and to have several key industries as a separate category. We classify firms into 16 industries: agriculture, forestry, and fishing (two-digit SIC code between 1-9), mining (10-14), construction (15-19), other manufacturing (20-27, 31-34), chemical (28-30), industry machine & transportation equipment (35, 37), electricity (36), transportation (40-47), communication (48), utility, i.e. electric, gas, and sanitary services (49), wholesale trade (50, 51), retail trade (52-59), depository (60), non-depository (61-69), service (70-79, 81-89), and health (80).

<Figure 5> presents ETR at the industry level for three groups of countries: all, OECD, and non-OECD. As in the investigation of over-time changes in ETR, we use a simple regression to examine ETR by industry.[5] Those figures are calculated using estimated coefficients of industry dummies in the regressions of ETR on industry dummies, period dummies, and country dummies. It is observed that on average, non-depository, service industries, and communication are less heavily taxed than construction, wholesale, depository, and utility. There are some differences between OECD and non-OECD countries. Agriculture and retail are more heavily taxed in non-OECD, while construction is more heavily taxed in OECD.

Estimating effective tax rates for individual sectors allows us to evaluate the cross-sector differences in effective tax rates within a country. We use the standard deviation of ETR across sectors within a country as a measure indicating a kind of industrial policy through taxation. The choice of the standard deviation as a proxy for cross-sector tax discrimination merits discussion. If a country decides to lowers (raises) tax rates for industries with rates higher (lower) than the average rate while preserving the average tax rates, it would represent a reduction in cross-sector discrimination and be well captured by a decrease in the standard deviation of ETR. On the other hand, a country reduces tax rates for each industry by the same %p, the level of ETR will be lower but standard deviation of ETR will not be changed.

The standard deviation of ETR is especially low in countries such as the U.S. (0.03), UK (0.06), Denmark (0.07), France (0.07), and Finland (0.07). Among OECD countries, Norway (0.16), Austria (0.15), and Belgium (0.15) are countries with large standard deviation of ETR. Note that these countries with higher standard deviation tend to have higher level of ETR.[6] Since we include both ETR and the standard deviation of ETR as independent variables in regressions, the standard deviation captures (not across industry) across country and period variability of ETR after controlling for the average level of ETR.

5. Preferential Tax Treatment and Industry Growth

Do preferential treatments in taxation promote growth in the favored industries? How do the preferential treatments affect capital formation and employment in the favored industries? How do industrial policy through taxation affect the overall growth of industry? This section provides empirical results to answer these questions.

Industry growth of the following five variables are used as dependent variable: after-tax earnings, before-tax earnings, net sales, total assets, and employment. The first two variables measure profits, the third variable measures output, and the last two variables measure inputs. Since corporate income taxes are basically on equity-financed capital, it is expected to affect assets much more strongly than employment. In the calculation of industrial growth, we do not drop firms with negative earnings, because we are measuring growth of industry not of firms. As mentioned earlier, we have four 5-year sub-samples. Combined with the classification of industries into 16 sectors, the maximum number of observation is 4,480 period-industry cells for 70 countries (=4*16*70) in total. Among 4,480 possible cells, we can estimate ETR and industrial growth for 1,486 cells. Dropping top and bottom 2% observations in the value of growth rate of before-tax profits leave us with 1,411 observations. We did so because there are some observations with extreme values of growth rates. In median regressions for the whole sample without dropping these observations with extreme values, we find qualitatively the same results.