“Empirical evidence on the impact of recent Korean tax reforms”

In 2011, Korea required all firms to report all value added tax (VAT) invoices electroni- cally to tax authorities. This unique law provided a natural experiment to examine the effects of this disclosure on income taxes and firms’ related responses. The authors find that this additional required disclosure caused firms to become less aggressive on their income taxes, and that they were unable to pass increased tax burdens forward to con- sumers or backward to suppliers and labor. To maintain, profitability firms cut research and development (R&D) costs, and this cost cutting was larger for tax aggressive firms. Policy implications of this unintended result are discussed.


INTRODUCTION
"Apple is thinking long-term: R&D spending keeps rising even as revenue dips", Business Insider, Jul. 26, 2016. Firms generally use research and development (R&D) spending to increase long-term value, not to meet short-term earnings targets (see for example, Almeida & Campello, 2007). Technology firms in particular tend to increase R&D spending each year under most conditions. For example, Apple spent USD 2.5 billion on research and development investment in the third quarter of fiscal year 2016, up to 25% from the prior year, despite the fact that the company continued to face revenue decreases 1 . Such increasing R&D spending is generally also true for Korean firms. However, Samsung Electronics, Apple's largest hardware competitor, reduced the number of its R&D centers from 44 to 41, and reduced its overall R&D spending by USD 45 billion in 2015 (the firm's first decrease since the Asian financial crisis in 1998). Such R&D cuts also occurred for other Korean firms around this period (starting in 2011), despite growth in these firms' product markets. A major purpose of this paper is to explain this unusual phenomenon in Korea as an unexpected by-product of tax legislation.
Firms' R&D investment decisions can be affected by various factors. One of the most important factors is the amount of surplus money the firm has, which affects all of its discretionary investment decisions ( Campello et al., 2010). These studies have generally found that firms facing financial distress are likely to decrease, or at least delay, current investment.
One way in which financially constrained firms have been found to generate additional funds for R&D (and other investments) is through tax avoidance behavior (Blaylock, 2016;Green & Kerr, 2016), despite the risk of being audited by tax authorities (Desai & Dharmapala, 2006;Rego, 2003;Slemrod, 2001; Ayers et al., 2011;Edwards et al., 2013;Brondolo, 2009;Campello et al., 2011Campello et al., , 2012 2 . Such increased R&D investment in turn may increase the firm's market value (Ayers et al., 2011). The above implies that when firms have less opportunities for aggressive tax avoidance or evasion behavior 3 , discretionary funds are lower and R&D investment should drop accordingly. Blouin et al. (2012) suggest that tax uncertainty is due to the inherent incompleteness of the law and its endless legislative, judicial, and administrative modification. Tax policy uncertainty also discourages investment (Niemann, 2011;Hassett & Metcalf, 1999;Croce et al., 2012) and may induce a reduction or delay investment (Stokey, 2013) 4 . A major reason for lowered tax avoidance is due to tax uncertainty. Blouin et al. (2012) found that tax unicertainty decreases aggressive tax avoidance, which in turn decreases firms' investments.
As discussed above, Korean firms exhibited significant R&D reductions starting in 2011, despite strong product growth. At this same time 5 , in an effort to curb tax evasion on value added taxes (VAT), the Korean government mandated that all invoices be electronic and that copies of such invoices be filed with the National Tax Service (Korea's equivalent of the US' Internal Revenue Service) 6 . Korea is the first country to have such an innovative system, and the Korean government reported that the result was an unqualified success in increasing VAT compliance. According to research conducted by the Korea Institute of Public Finance in 2011, the system saved about KRW 900 billion (approximately USD 790 million) in compliance costs (Korean NTS Annual Report, 2015). Such electronic VAT invoices were required to be filed by the 10th day of the month following the month of issuance; a penalty tax (1% of the VAT-invoiced amount) was levied on companies that either did not file or were late in filing. The penalty tax was either 0.5% or 1%, depending upon the number of days the filing was delayed. To incentivize use of the system, suppliers receive a tax credit of 100 KRW (slightly under 9 cents US) per issuance, and up to KRW 1 million can be credited per year per taxpayer. The NTS reported that by July 31, 2011, 99% of all eligible corporate taxpayers participated in issuing electronic invoices for VAT, amounting to KRW 1.142 trillion (99.9% of all invoiced amounts) 7 .
This natural experiment provides the opportunity to examine a number of other potential outcomes. Because VAT invoices provide a more transparent (and immediate) trail for income tax audits, the new law may have reduced income tax evasion and avoidance as well 8 . We find that this in fact occurred. We next test whether firms were successful in passing such additional tax burdens onto consumers in the 2 Such firms are also more likely to use tax avoidance if they are not able to opportunistically achieve earnings targets other than taxes (Dhaliwal, Gleason, & Mills, 2004). 3 Since either avoidance or evasion have the same objective here to increase funds for R&D, we can argue that they are similar. 4 Taxes are a particularly important matter for multinationals' investment decisions, and such firms' tax decisions are based on the average effective tax rate. See "How Taxes Affect Investment Decisions for Multinational Firms?" Forbes, Apr. 14, 2015. 5 This was preceded in January 2001 by a law change whereby suppliers were permitted to issue electronic VAT receipts. Effective July 1, 2007, the self-billing of VAT invoices was encouraged to improve the transparency of VAT transactions when a businessperson receives a supply of goods or services and is not able to take receipt of tax invoices. However, the supplier of goods or services could not issue a tax invoice to avoid exposing the tax base. Self-billing was permitted if the head of a district tax office approved, and as long as the consideration of relevant transactions was more than KRW 100,000 and less than KRW 5,000,000. 6 Earlier law changes allowed such invoices, but there was limited adoption. Effective January 1, 2001, suppliers were permitted to issue electronic VAT receipts. Effective July 1, 2007, the self-billing of VAT invoices was encouraged to improve the transparency of VAT transactions when a businessperson receives a supply of goods or services and is not able to take receipt of tax invoices. However, the supplier of goods or services may not issue a tax invoice to avoid exposing the tax base. Self-billing was performed upon the approval of the head of a district tax office, as long as the consideration of relevant transactions is more than KRW 100,000 and less than KRW 5,000,000. 7 Annual Report 2011, Korean National Tax Service. 8 The main goal of e-invoicing was to reduce tax evasion. We argue that it also reduced tax avoidance.
form of higher prices, or backward to labor (via lower wages) or to vendors (via lower prices). We find that it was not the case, which implies that the increased tax burden may have resulted in lower after-tax profitability. These findings have important policy implications for countries adopting similar systems; with increasing use of ERP systems by companies and the ability to transmit such data real-time over the web, such systems may not only be feasible and cost-effective, but also reduce tax evasion with no "side effects" on other sectors of the economy.
Finally, we find that in order to maintain profit margins, firms reduced R&D expenditures. Such R&D reductions were more concentrated in firms, which were prone to tax avoidance (aggressive tax behavior), implying that such firms were most affected by the policy. This latter finding is consistent with the "flip" of findings in previous studies (e.g., Ayers et al., 2011). and in doing so, adds to the literature in this area; that is, firms, which are more tax aggressive, tend to spend more on R&D. The R&D results have policy implications for all countries insofar as tax increases (resulting from decreased avoidance opportunities) may inadvertently result in decreased innovation by firms. The lower profitability and decreased tax avoidance results have policy implications for other countries considering similar electronic VAT reporting systems. The remainder of the paper is organized as follows. Section 1 provides a literature review and hypothesis development. Section 2 discusses the sample and methodology. Section 3 presents descriptive statistics, correlations, and regression results. Final section discusses the results and related policy implications. 9 Firms also use tax avoidance to achieve financial earnings targets (Dhaliwal, Gleason, & Mills, 2004).

LITERATURE REVIEW AND HYPOTHESES DEVELOPMENT
As noted in the previous section, Korean firms significantly cut R&D after the 2011 tax reforms. Before examining the tax law itself, it is instructive to examine the literature on firm discretionary investment, including investments in R&D. Internal funds are significant forecasters of firms' investments (Almeida & Campello, 2007;Fazzari et al., 1988;Lamont, 1997 Stokey (2013) analyzes whether uncertainty about future tax policy leads to a temporary reduction in investment and finds that firms delay projects until the uncertainty is resolved.
The conclusions from the above literature are that investments in R&D are a function of cash flows, and that such cash flows are in part generated by aggressive tax behavior (tax avoidance). But such tax avoidance (and therefore investments) are reduced where there is uncertainty. In the case of the Korean tax law changes in 2011, tax avoidance opportunities are decreased, since firms must fully disclose all transactions to the government. Additionally, there is uncertainty related to government audit likelihoods from the new law due to the law's "newness". Both of these reductions in tax avoidance reduce cash flows and, accordingly, reduce investments in R&D.
Based on the aforementioned literature, we posit: H1: Firms' R&D investments are positively associated with firms' tax avoidance 10 .
H2: Firms' R&D investments are negatively associated with firms' tax avoidance after tax compliance systems are implemented.
We additionally test if companies were able to shift taxes. Since the effects of the law were economy-wide, it is conceivable that this caused shifts in demand and supply curves in numerous markets. The general equilibrium effects of this are described in Harberger (1962) and a plethora of subsequent papers (see Gravelle, 2010 for a survey of the related tax incidence literature). The net effects could conceivably be that corporations shift some of the extra taxes forward to consumers in the form of higher prices, and/or backward to labor in the form of lower wages or to vendors in the form of lower prices paid. Any taxes not absorbed would result in lower after-tax profits, which is essentially a tax on capital. The magnitudes of such shifting depend on elasticities in all markets and are essentially an empirical question. It is also conceivable that because all firms (corporate and non-corporate) were affected by the law, opposing pressures on prices between vendors and buyers could have been offsetting. Similarly, it may be there were no effects in labor markets since there could have been equal shifting supply and demand between sectors.

Direct effects on firms
Although the intent of the law change was explicitly to reduce VAT evasion, the additional disclosure of all transactions to tax authorities may also have reduced income tax evasion and avoidance 12 . This could occur in a number of ways. For example, firms previously wishing to avoid tax could avoid issuing some receipts, but this was no longer feasible under the new law. Firms wishing to accelerate deductions or delay receipts could no longer do so. Additionally, since inter-company receipts were also to be reported, firms utilizing tax aggressive transfer pricing schemes would now have the prices completely exposed to tax authorities 13 .
Conceivably, this could make firms less prone to utilize such methods. To test whether tax avoidance decreased, we examine book-tax differences before and after the law change in the following regression model: YEAR are annual indicator variables. Tax avoidance measurements are based on book-tax differences, or BTDs. The BTD is the difference between the accounting income before tax and TI is the taxable income, and it is scaled by total assets. Taxable income is calculated as Income Tax Expense + {(Deferred Tax Assets t -Deferred Tax Assets t-1 ) -(Deferred Tax Liabilities t -Deferred Tax Liabilities t-1 )}. Thereafter, it is divided by the tax rate. Following the method developed by Desai and Dharmapala (2006), an OLS regression model is used to account for the component of the booktax gap, which is attributable to earnings management, as follows: where TA is total accruals calculated by subtracting operating cash flows from net income using the measure of total accruals developed by Hribar and Collins (2002), which is divided by the begin- ning of year assets. The residual is then BTD after accruals, and it is our measure of TAvoid in equation (1).
The model in equation (1) includes control variables that can affect tax avoidance. These variables include leverage, size, operating cash flow, return on assets, deferred tax, loss in the prior year, sales growth, investment, foreign shareholders, and auditors (Big 4 vs. non-Big 4). Leverage is the total liabilities divided by total assets. Leverage may have a positive association with tax avoidance due to tax-deductible interest payments. Size, which is measured as the natural log of total assets, is included to control for size effects. Size may have a positive association based on political cost theory (Watts & Zimmerman, 1986) or negative association based on political power theory (Siegfried, 1974) with tax avoidance. Since tax payments are a significant cash outflows, operating cash flow is included. Return on assets, which is measured as net income divided by total assets, is included to control for firm profitability 14 .
Deferred tax is included to control for future variable tax expenses. It is expected that the data will show that firms that reported losses in one year are more likely to be tax avoidant in the subsequent year. Thus, a LOSS dummy variable is included. Growth firms are more likely to purchase tax-favored assets (Chen et al., 2010), so sales growth is included to control for growth. Since firm's investment decisions might have an effect on tax planning, investment is used as a control variable. Foreign shareholders and auditors (Big 4 vs. non-Big 4) are also included, since they may have an impact on reducing tax avoidance through monitoring and oversight.

Collateral effects: R&D expenditures
The following regression allows for tests for both firms' post-law change cuts in R&D, and for the incrementally higher effect of this for firms with higher tax avoidance:  14 Gupta and Newberry (1997) and Rice (1992) found a positive association between firm profitability and tax avoidance.
where & R Dintensity is the ratio of R&D expenditures to firm sales (e.g., Greve, 2003 TaxAvoidance is tax avoidance as defined previously, TaxSystem is a dummy variable for the tax compliance system, (also defined previously), The other factors and control variables include size, financial distress, operating cash flow, return on equity, Tobin's q, market-to-book ratio, sales growth, foreign shareholders, operating cycle. Size, which is measured as the natural log of total assets, is included to control for size effects on R&D investment (Shehata, 1991). Size may have a positive association with R&D investment. Financial distress is measured following the modified Altman Z-score (Graham et al., 1998). This model is calculated as Z = 1.2 (working capital divided by total assets) + 1.4 (retained earnings divided by total assets) + 3.3 (earnings before interest and taxes divided by total assets) + 0.999 (sales divided by to.tal assets). Financial distress may have a positive association with R&D investment, as firms facing financial distress have less funds for discretionary investment. Since firms with greater cash flows are more likely to invest (Stein, 2003), operating cash flow is included. Return on equity, which is measured as net income divided by equity, is included to control for firm profitability. More profitable firms are likely to have funds available for investment, we thus expect positive sign. We also include Tobin's q (1969) and market-to-book ratio to control for investment opportunities. Tobin's q is calculated as the market value of equity plus liabilities, divided by total assets. Sales growth is included to control for growth and to capture changes in demand that would require additional investment (Bond et al., 2007). We expect a positive association of sales growth with R&D investment. Foreign shareholders are also included, since they may have an impact on controlling R&D investment through monitoring and oversight. We also include a control for operating cycles (opcycle), as firms may have different investment needs depending on their life cycle (Dechow, 1994;Dechow, Kothari, & Watts 1998;Dechow & Dichev, 2002). Opcycle is calculated as the natural log of the sum of receivables turnover (RECT/SALE) and inventory turnover (COGS/ INVT) multiplied by 360. Notes: TAvoid: book-tax gap residual calculated using the method developed by Desai and Dharmapala (2006), RD: total R&D investment divided by sales, Post2011: coded 1 for the period after 2011, otherwise 0, LEV: total liabilities divided by total assets, FD: financial distress measured following modified Altman Z-score, SIZE: the natural logarithm of total assets, OCF: operating cash flow divided by total assets, ROA: net income divided by total assets, ROE: net income divided by owners' equity, DT: deferred taxes divided by total assets, LOSS: coded 1 if the firm has incurred a loss in the previous year, and 0 otherwise, TQ: Tobin's q, calculated as the market value of equity plus liabilities, all divided by total assets, MTB: market to book ratio, market value of equity divided by book value of equity, FSH: foreign investor ownership, GROW: sales growth, INV: plant, property, and equipment (except land and construction in progress) divided by total assets, OPcycle: the natural log of the sum of receivables turnover and inventory turnover multiplied by 360, BIG4: coded 1 if the auditor is a Big 4 auditor, and 0 otherwise.

Descriptive statistics and correlations
OCF, ROA, DT, LOSS, INV) (p < 0.01). Significant negative correlations are visible between tax avoidance and some other control variables (LEV, BIG) (p < 0.01). Significant negative correlations are observed between R&D investment and Period 2011 interaction term with high BTD dummy, which is coded as 1 if tax avoidance is greater that the median (p < 0.01). Significant positive correlations are also seen between R&D investment and some control variables (SIZE, OCF, ROA, DT, LOSS, INV) (p < 0.01). Significant negative correlations are visible between tax avoidance and some other control variables (LEV, BIG) (p < 0.01). Variance Inflation Factors (VIFs) were calculated and indicated that no multi-collinearity problems are evident.

Regression results
Panel A of Table 4 reports OLS regression results for the association between the enforcement of tax compliance systems and tax avoidance. As can be seen, book-tax differences decreased (significant at 0.01), a result consistent with the prediction that the law had a collateral effect on income tax evasion and avoidance. The regression results are supported by means in Panel B, which show that taxable incomes increased and book-tax differences decreased after the law was enacted (both results significant at 0.001).   TAvoid Post2011 LEV  SIZE  OCF  ROA  DT  LOSS  FSH GROW INV  BIG4   TAvoid  1  --------- Note: See Table 1 for variable definitions, * significant at 5% level.
variable is sales/COGS, there was no increase in prices after 2010. Similarly, when the dependent variable is sales/total assets, there was no indication of forward shifting. On the other hand, when the dependent variable is consumer prices, there is a 22% decrease in consumer prices 16 . The net results suggest that any increases in income taxes were not passed forward to consumers.
Panel B reports regressions looking at tax shifting backward to labor and/or vendors. The regression where the dependent variable is COGS/total assets shows no shifting backward to inventory (and other) vendors, or to labor as part of COGS, after 2010. The regression where the dependent variable is wages/ total assets shows that there was no backward shifting of taxes in the form of lower wages after 2010. Finally, the regression having the dependent variable as SG&A expense/total assets shows a statistically significant drop. As discussed below, SG&A includes discretionary expenditures such as advertising and R&D. The potential decrease in R&D is examined next. 16 The data are from the Consumer Price Index reported by the Economic Statistical System, Bank of Korea (various years).
A telling finding is shown in the first column in Panel A. Here, the dependent variable is after-tax ROA. Results show no change after the tax law change. Thus, firms were able to maintain their profitability despite tax increases and despite the fact they were unable to shift taxes forward or backward. This suggests that they had to have cut costs somewhere, which is more closely examined in the next section. Table 6 shows regression results for R&D intensity. The BTD variable, which covers all periods, is positive and significant at 0.01, a result consistent with prior research. That is, firms prone to tax avoidance use the resultant cash flows for R&D. The interaction term of BTD and post law-change is negative and significant at 0.01. Thus, firms prone to tax avoidance decreased their R&D after the law change. This finding is consistent with the prediction that it was more difficult after 2010 for firms to avoid taxes, and the resultant decrease in funds left less for R&D investment. Note: * p < 0.10, *** p < 0.01. Note: t-values are shown in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. Notes: HBTD2011: Period 2011 interaction term with HBTD dummy, HBTD dummy is coded as 1 if tax avoidance is greater that the median. Otherwise, it is coded as 0. Other variables: see Table 2 for variable definitions, t-values are shown in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.

CONCLUSION
In 2011, Korea required all firms to report all value added tax (VAT) invoices electronically to tax authorities. This unique law provided a natural experiment to examine the effects of this disclosure on income taxes and firms' related responses. We find that this additional required disclosure caused firms to become less aggressive on their income taxes, and that they were unable to pass increased tax burdens forward to consumers or backward to suppliers and labor. To maintain, profitability firms cut research and development (R&D) costs, and this cost cutting was larger for tax aggressive firms.
There are a number of policy implications of this study. Although the Korean law change is unique, it is conceivable that other countries will follow its lead in order to reduce tax evasion for VAT purposes. With the increasing use of ERPs by firms and the ability to transmit such data real-time over the web, such a system is feasible in many industrialized countries. If other countries follow suit, VAT tax evasion will be dramatically reduced, and there may be minimal collateral effects on market prices throughout the economy (i.e., distortionary effects) since firms will absorb the extra tax costs and not pass them forward to consumers or backward to supplier and/or labor. Countries adopting the Korean system may also be able to reduce income tax evasion/avoidance due to the extra disclosure requirements for transactions, especially those affecting transfer pricing.
On the other hand, a major negative effect could be a reduction in discretionary expenditures such as in R&D, since firms would have less funds from tax avoidance and evasion to invest in R&D. Such decreases in R&D could result in lower innovation and thus be a non-trivial welfare cost to that country's economy to the extent the country is innovation-driven. Other countries adopting such a system should therefor consider the tradeoffs of increased tax revenues versus the potential of slowed economic growth due to decreased investments in R&D.