Article Type: Original Article Do Political Connections Weaken Tax Enforcement Effectiveness?* Kenny Z. Lin, Lillian F. Mills,† Fang Zhang, and Yongbo Li September 2017 Kenny Z. Lin Department of Accountancy, Lingnan University, Hong Kong Email: firstname.lastname@example.org Lillian F. Mills McCombs School of Business, University of Texas at Austin, Austin, TX 78712 Email: Lillian.email@example.com Fang Zhang Department of Accountancy and Law, Hong Kong Baptist University, Hong Kong Email: firstname.lastname@example.org Yongbo Li Department of Accounting, Hong Kong University of Science and Technology, Hong Kong Email: email@example.com * Accepted by Jeffrey Pittman. We appreciate comments from the Arizona Tax Readings Group, 2015 EAA Conference, Wei Cui, Lisa De Simone, Michael Devereux, Michael Donohoe (discussant at the 2014 AAA Annual Meeting), Katharine Drake, Matthew Erickson, Nathan Goldman, Jeffrey Gramlich, Jeffrey Hoopes, Kenneth Klassen, Colin Koutney, Stacie Laplante, Kelvin Law, Georgia Maffini, Edmund Outslay (discussant at the 2015 NTA Annual Conference), Jeffrey Pittman, Braden Williams, Jing Xing, Li Xu, and students and faculty at The University of Paderborn, Washington State University, Xiamen University, and Zhongnan University of Economics and Law. † Corresponding author: Lillian.firstname.lastname@example.org This is an Accepted Article that has been peer-reviewed and approved for publication in the Contemporary Accounting Research, but has yet to undergo copy-editing and proof correction. Please cite this article as an “Accepted Article”; doi: 10.1111/1911-3846.12360 This article is protected by copyright. All rights reserved. Do Political Connections Weaken Tax Enforcement Effectiveness? ABSTRACT This paper examines whether ties to politicians by corporate boards of directors weaken the effectiveness of tax authorities in constraining tax avoidance in China. We use a unique data set to measure geographic time-variant tax enforcement, including the probability of income tax audit, the expertise of tax officers, and the consequences of underreporting tax liabilities. Based on a sample of 11,121 firm-years from 2003 to 2013, we find that the deterrent effect of the probability that a firm’s taxable income understatement will be detected and lead to heavy penalties is significantly undermined if the board is politically connected. To enhance our analysis, we use opportunities for income shifting, the most likely mechanism through which Chinese firms avoid taxes on an ongoing basis, to illustrate how connected boards exert power to unwind the constraining effect of tax enforcement. Overall, our results suggest that a board’s ties to politicians can be a significant challenge to the effective enforcement of tax compliance in a politically controlled economy. Keywords Effective tax rates, income shifting, political connections, tax avoidance, tax enforcement. JEL Descriptors H25 H26 M40 This draft: September 2017 Do Political Connections Weaken Tax Enforcement Effectiveness? 1. Introduction We investigate whether politicians tied to a firm’s board wield sufficient power to mitigate or even completely offset tax enforcement effectiveness. Prior research has taken a piecemeal approach to examining the association between tax authority enforcement and tax avoidance (Hoopes et al. 2012) and between corporate political connections and tax aggressiveness (e.g., Brown et al. 2015; Kim and Zhang 2016). However, few studies to date have examined whether less scrutiny from revenue authorities of the aggressive tax positions of politically connected firms is also a reason for corporate tax avoidance, or whether government enforcement is less effective for firms with political connections.1 Kim and Zhang (2016) 1 Desai et al. (2007) find a sudden increase in tax enforcement (i.e., increased tax payments, decreased related party trades, and reduced tax haven entities) in Russia following the election of Vladimir Putin as prime This article is protected by copyright. All rights reserved. call for research to examine the interplay of political ties and tax enforcement. We respond to this call directly by examining whether board of director ties to politicians, a widespread phenomenon in China, limit the ability of revenue enforcers to constrain tax avoidance. We also answer Crocker and Slemrod’s (2005) call to extend the relatively sparse literature on the effects of tax authority monitoring on corporate tax behavior (El Ghoul et al. 2011; Guedhami and Pittman 2008; Hoopes et al. 2012). Our question deserves particular attention because tax enforcement affects government revenues available to fund public expenditures. By studying the political economy of tax enforcement, we push the boundaries of prior research that separately examines whether political connections facilitate tax avoidance while enforcement efforts impede it, with little regard to the possibility that the process of enforcing tax laws is itself prone to political influence. We measure tax avoidance with the one-year book effective tax rate (ETR) that equals the ratio of total income taxes to pre-tax income before special items (Hoopes et al. 2012). We use a novel data set sourced from the China Tax Audits Yearbook, published annually by the State Administration of Taxation (SAT, akin to the IRS), to measure aggregate tax enforcement characteristics for all provinces and large cities in China. We use detailed and summary measures of enforcement for each region-year: the probability of tax audits, the expertise of tax officers, and the consequences of underreporting taxable income. We consider both the existence and intensity of a board’s political connections, namely whether the chairman is connected and the number and percentage of connected board directors. To improve identification, we also consider a change in the political connection status from not minister in 2000. Chen et al. (2015) examine three channels through which the political contributions of firms affect their tax burden in the United States and find evidence consistent with the influence of political contributions on IRS audit rates and hence on the tax burden through enforcement of tax laws and regulations. However, Chinese firms do not establish political ties through political campaign contributions or corporate lobbying expenditures but through hiring board members with political backgrounds. The current organizational structure of Chinese tax administration places local tax offices at risk of intervention by local politicians. This article is protected by copyright. All rights reserved. connected to connected. We predict that firms with a well-connected board are subject to less scrutiny so that enforcement of tax avoiders is less effective in the presence of a politically connected board. We report the following findings based on a sample of 11,121 firm-years from 2003 to 2013. First, corporate political connections attenuate the constraining effect of tax authorities on tax avoidance in pooled cross-sectional tests. Holding various firm characteristics constant, the existence of political connections on the board reduces the sensitivity of ETR to tax enforcement by approximately 86 percent to 88 percent, depending on model specifications. To strengthen the causal inference, we then analyze the effect of a connection status change triggered by firm-specific and market-wide factors. Specifically, we test whether the incremental association between ETR and Enforcement is weaker for firms with an increase in connected directors than for firms without a similar increase. We also use the 2012 nationwide anti-corruption campaign as a natural experiment to implement the difference-in-differences (DID) analysis. If connected firms have the ability to obtain or maintain lenient regulatory treatment, then, arguably, firms that lose a connection following the crackdown on corrupt officials will experience a drop in this ability. Using a DID procedure, we find that the positive ETR-Enforcement relation weakens after a firm acquired political ties, but strengthens after a change in the political landscape that diminished the rent-seeking and favor-exchanging ability of both politicians and managers. Triangulating our cross-sectional and time-series differences helps confirm that our results are not due to common shocks or omitted variables. In supplemental tests, we find that our results are robust to measuring tax avoidance using a cash ETR or a three-year book ETR adjusted for firm size and industry effects (Balakrishnan et al. 2014; Kim and Zhang 2016). To help us better understand the underlying mechanism, we also examine the most common opportunity for tax avoidance in China: the This article is protected by copyright. All rights reserved. shifting of income between member firms in a business group (Lo et al. 2010; Shevlin et al. 2012). China requires consolidated financial reporting but separate legal entity tax filing. Such an arrangement, similar to the separate reporting under the territorial system (Markle 2016; Altshuler and Grubert 2001), creates strong shifting incentives. We model the opportunity for such shifting as a joint function of the spread in tax rates among members of the consolidated group and the magnitude of intragroup transactions, like Jacob (1996). Based on a reduced sample of 7,869 group firm-years from 2003 to 2013, we find that tax avoidance associated with income-shifting opportunities is constrained by tax enforcement but unwound when firms are politically connected. China provides a good setting for conducting this research. First, despite its rapid transition to a market-based economy, China remains a hierarchical, relationship-based society, with political connections (an important form of relationship) being critical to firm survival and success (Xin and Pearce 1996). Second, the current arrangement that local tax authorities are under dual supervision from central and local governments exposes them to a great risk of interference by local politicians. Third, while the country has uniform tax laws, enforcement often deviates from written law in some regions (PricewaterhouseCoopers 2012). Our study also provides several enhancements in research design. First, in sharp contrast to prior non-U.S. research that suffers from not having data on when political connections are formed or severed (e.g., Chaney et al. 2011; Guedhami et al. 2014), we improve identification by exploiting changes in the political connection status. In an attempt to alleviate the concern that endogeneity biases our main findings, we also take a number of steps, including DID analysis, a Heckman two-step treatment effects procedure, and a propensity score matching approach. Second, our tax enforcement proxies benefit from varying on multiple dimensions in comparison to extensive cross-country research that relies on proxies that only vary on a single dimension (e.g., Dyck and Zingales 2004; Haw et al. This article is protected by copyright. All rights reserved. 2004; Desai et al. 2007). Finally, the various institutional characteristics (e.g., extensive government involvement in business, varying statutory tax rates across industries, separate legal entity tax filing, and widespread intragroup transactions) that we describe later enable us to examine the research question in a high-power testing ground. Although our single-country focus inevitably undermines external validity of our results, our description of informal institutions and ineffective law enforcement, which are common in the BRIC and emerging economies (Estrin and Prevezer 2011; Faccio 2006; Khwaja and Mian 2005), helps dispel this validity concern. This study contributes to the very limited literature exploring the interaction between politics and tax administration receptive to bureaucratic influence. Our empirical results raise questions about the effectiveness of the deterrent role of tax authorities and highlight the importance of understanding the political economy of tax enforcers in a transition economy. From a policy viewpoint, the results suggest that reforms to increase government revenue should not be limited to the strengthening of tax administration within the revenue system alone, but should also take the political economy of tax enforcers into account. 2. Tax aggressiveness and tax enforcement in China The most common methods by which Chinese firms avoid income tax include underreporting sales revenue, fabricating and inflating expenses, abusing tax credits, using unsanctioned different policies for book and tax purposes, managing discretionary accruals, and transferring profits (across periods within the same firm or across member firms in a business group) (Cai and Liu 2009; Lin et al. 2012; Lin et al. 2014).2 As the first four methods represent tax evasion, it is unlikely that firms can apply these methods for an extended period. In contrast, the last two could represent legitimate tax planning that allows managers 2 According to Benari (2016), tax avoidance schemes in the United States fall into three broad categories: “schemes involving financial transactions (e.g., debt buying, overpayment, use of dividends, deductions on employee stock options); schemes involving geographical locations (e.g., tax havens, outward domestication); and schemes involving cost of goods (e.g., false invoicing, transfer pricing).” Comparatively, outbound investments, overseas tax havens, and deductions of employee stock options are less applicable in China. This article is protected by copyright. All rights reserved. to shift earnings via accounting accruals or real activities for long periods. Moreover, tax authorities usually find it difficult to detect and unravel income shifting, because it is subject to managerial choices that are difficult for external parties to challenge. Therefore, accruals management and income shifting are more prevalent than other methods of avoiding income tax over the long run in China. In an effort to attract investment to specific regions and industries, China offers preferential tax rates (typically 15 percent) to domestic firms operating in the designated zones and industries. As a result, Chinese statutory tax rates have wide cross-sectional/regional and temporal variation, similar to international variation. A Chinese parent firm must consolidate the income of its subsidiaries for book but file separate tax returns on a legal entity basis.3 Therefore, the parent firm can reallocate taxable income within the group without affecting consolidated pre-tax book income. Chinese group firms usually have a large, powerful parent (e.g., a listed firm) whose subsidiaries and affiliates operate in different regions and industries subject to varying tax rates. The parent firm exercises its authority over subsidiaries through multiple layers to achieve various objectives (Fan et al. 2007). For example, it can strategically transfer income within the pyramid to reduce the tax burden for the group (Lo et al. 2010; Shevlin et al. 2012). Transfer pricing, also referred to as base erosion and profit sharing, is a major tool for corporate tax misreporting, 4 because applying the arm’s length principle is challenging particularly for an emerging economy whose tax administration is handicapped by 3 4 Separate taxation is practiced in developed (e.g., Canada, Norway, Sweden, and the United Kingdom) and developing countries (e.g., Brazil, Greece, India, Korea, Russia [for small groups], and South Africa) (taxsummaries.pwc.com/). In September 2014, the OECD approved the Base Erosion and Profit Sharing (BEPS) Action Plan to create a single set of international tax rules to minimize the erosion of tax bases and the artificial shifting of profits for tax purposes. The OECD issued its report, Measuring and Monitoring BEPS, in October 2015, and China plans to localize the BEPS recommendations (PricewaterhouseCoopers 2015). Transfer pricing issues have led to some of the largest U.S. tax settlements, and firms filing an IRS Schedule UTP (uncertain tax position) must disclose whether the position involves transfer pricing (Towery 2017). Lin et al. (2012), Lin et al. (2014), Lo et al. (2010), and Shevlin et al. (2012) document that Chinese firms avoid subnational tax through domestic income shifting. This article is protected by copyright. All rights reserved. underdeveloped financial markets and inadequate financial resources and human expertise (Dharmapala and Hines 2009; He et al. 2012; Piotroski and Wong 2012). In this institutional setting (i.e., differential tax rates, separate tax reporting, a multilayered pyramid structure, and insufficient market-supporting infrastructure), one can expect tax avoidance through income shifting to be rampant in China. China’s revenue agency has been stepping up efforts to challenge tax avoidance via transfer pricing both inside China and in cross-border transactions since 2000. Transfer pricing audits frequently result in tax deficiencies with proposed adjustment, interest levy, and penalties.5 The high ex ante threat of a tax audit, heavy penalties, and ex post incidence of closer monitoring should deter firms from taking aggressive tax positions. Recent evidence suggests that tax authority oversight can constrain opportunistic managerial reporting. 6 However, there is anecdotal evidence that casts doubt on the capability or willingness of China’s tax administration to detect and constrain sophisticated tax avoidance arrangements. As China decentralized decision rights, local governments acquired more autonomy in economic matters (Li and Zhou 2005) and as a result had goal incongruence with the central government. The Chinese saying “the mountains are high and the emperor far away” aptly captures the autonomy that is abused by some local officials. The decentralization also created a wide disparity in economic and institutional development across regions (Fan et al. 2011). Consequently, there are differences in government involvement, professionalism of business intermediaries, and the quality of local government 5 6 For example, in 2014, Xuzhou SAT of Jiangsu province claimed that an agreement for sharing management and R&D expenses lacked economic substance and represented abusive transfers of profit between two differently taxed affiliates. These two firms had to pay an extra tax charge of RMB 5 million (Huang and Zhao 2015). Adjustments are subject to a special interest levy based on the RMB loan base rate applicable to the relevant period of tax delinquency plus five percentage points. A daily surcharge of 0.05 percent is levied on the underpaid taxes and interest levies. What is worse, tax agents will follow up closely for five years after any adjustment, during which period post-adjustment firms must submit contemporaneous transfer pricing documentation by June 20 of each year. For example, active enforcement of tax laws results in lower corporate tax avoidance (Atwood et al. 2012; Hoopes et al. 2012), reduces rent diversion by managers (Desai et al. 2007), improves the quality of financial reporting (Guenther and Young 2000; Hanlon et al. 2014), and lowers the cost of debt and implied equity capital (El Ghoul et al. 2011; Guedhami and Pittman 2008; Houston et al. 2014). This article is protected by copyright. All rights reserved. (World Bank 2006). Government quality influences the administrative effectiveness of local tax bureaus. Although tax laws are centrally legislated, local tax offices may interpret and enforce them differently (PricewaterhouseCoopers 2012). In some areas, tax officials interpret tax laws and the arm’s length principle more strictly and take a tougher stance against noncompliance, while in other areas they are more lenient in imposing penalties, fines, and surcharges. As a result of uneven economic growth across regions, some tax offices also have more financial resources than others to attract and retain top talent, enforce tax laws, improve tax administration infrastructure, and advocate for voluntary tax compliance as a form of corporate social responsibility. 3. Research hypothesis: Political connections constrain tax enforcement effectiveness Since the 1978 reforms, China has developed a unique hybrid economy melding socialist and capitalistic ideas by balancing government ownership of firms with the rapidly growing private sector. However, China remains a hierarchical, relationship-based society, with political connections being an important form of relationship. The government still has absolute power over the allocation of massive state resources (e.g., licenses and permits, initial and subsequent public offerings, bank loans, subsidies, and government contracts) and effective control of large-scale state-owned enterprises (SOEs) and major banks, which continue to dominate key sectors of the economy. The Communist Party has ruled the country since 1949, without either challenge by opposition or scrutiny by the media. In the absence of political reform, corporate ties with bureaucrats have become even more important to take advantage of the many new business opportunities created by the economic reform. To compete with SOEs for economic resources and reduce the risk of expropriation, many Chinese private firms hire managers who are government officials or current members This article is protected by copyright. All rights reserved. of the People’s Congresses or Political Consultative Conference (Li et al. 2006).7 Unlike many mature markets that impose reasonable restrictions on a civil servant’s rights to work after leaving office, China has no specific post-service employment rules for civil servants. Moses (2015) estimates 2 percent of the bureaucratic workforce left government for better opportunities in the yearly 2000s. Thus, many firms could hire former government officials and tax experts to act as a protective shield against fines, litigation, loss of explicit and implicit contracts, negative publicity, and damaged relations with the government.8 China has a two-tier tax administration system of national and local tax bureaus. The SAT is the highest tax authority. The tax organizations at and below the provincial level are divided into the offices of the SAT and local tax bureaus, with the former (latter) collecting corporate income taxes from centrally (locally) controlled enterprises. Local tax bureaus are under dual leadership from central and local governments, but the local government determines director appointments, organizational structure and size, staffing levels and cadre promotions, and operational budgets and employee pay scales. Such arrangements place local tax agencies at risk of interference by local politicians. Both anecdotal and empirical evidence suggests that Chinese tax administration is subject to political pressure. 9 For example, Liu et al. (2015) find that political leaders tend to tighten collection to generate 7 8 9 The People’s Congresses or Political Consultative Conference have a significant influence on and a close relation with government officials. Many members of these organizations are current or former key government officials. Although the membership of the two organizations includes non-Communists (e.g., movie superstar Jackie Chan), all government officials must be Communist Party members. The Chinese Communist Party reached almost 88 million members in 2014 (South China Morning Post, June 30, 2016), representing 6 percent of the Chinese population. Party members include factory workers, nurses, and teachers, so party membership alone does not equate to the political power to influence corporate affairs. Cases attracting media attention include the governor of Shandong province and the deputy head of the SAT joining Sinotruk Hong Kong Ltd., the vice-mayor of Xinghua city in Jiangsu province working in China Pacific Construction Group Ltd. while still in office, senior officials at the national securities regulatory authority joining the private sector, and anti-tax avoidance experts from the SAT joining leading international accounting firms as well as public and private corporations. A local government can affect the operation of local tax bureaus via its finance bureau, which controls resource allocation and staff welfare subsidies to all bureaucratic agencies. A change in government officials can change tax enforcement behavior. For instance, when Xilai Bo was removed from the Central Committee of the Communist Party of China, on March 15, 2012, the Chongqing municipal government immediately instructed its Bureau of Finance to heighten tax enforcement (21st Century Business Herald, February 1, 2013). This article is protected by copyright. All rights reserved. immediate effects on fiscal revenue during the early part of their terms. Chen et al. (2015) also find that newly appointed local leaders are inclined to collect more taxes to expand fiscal expenditure. Chen (2014) shows that under the influence of the local Communist Party secretary, local tax authorities impose lax enforcement of value-added tax collection on capital-intensive industries that participate in corruption. Deng and Luo (2011) provide evidence that when bureaucrats are evaluated on collections, local tax departments collude with government-friendly businesses by timing tax collections to achieve their annual tax budget. Tax auditors are also likely motivated and shaped by private interests. Families of local tax officials face welfare and social issues, such as housing, schooling, education, healthcare, and employment, the resolution of which depends, to a large extent, on the ability of bureaucrats to cultivate a wide social network of contacts with firm managers. Chinese tax agents may subject the tax positions of favored or connected firms to less scrutiny or arbitrarily enforce tax regulations to promote these firms over others (Shevlin et al. 2012). Tax and non-tax research suggests that political connections help individual firms, particularly in economies where the government exercises tight control over resource allocation. Such value includes government bailouts in the event of financial distress (Faccio 2006), increased firm value (Fisman 2001; Johnson and Mitton 2003), better access to the legal system and bank loans (Khwaja and Mian 2015), preferential tax treatment (Adhikari et al. 2006; Faccio 2006), and lower regulatory penalties and relaxed government oversight (Correia 2014; Wu et al. 2016). Desai et al. (2007) model a political economy game that involves the state, insiders, and outside shareholders in the determination of tax liabilities, in which managers may bribe the tax authority to make it an accomplice to their diversion. The U.S.-based results suggest that politically connected firms can shape IRS audit intensity in their favor through influencing the allocation of enforcement-related resources and execution This article is protected by copyright. All rights reserved. of tax laws and regulations (Bagchi 2012; Chen et al. 2015; Young et al. 2001). Connected firms face lower penalties if they are prosecuted by the SEC (Correia 2014) and accrue greater future tax benefits (Brown et al. 2015; Francis et al. 2012; Kim and Zhang 2016). Kim and Zhang (2016) call for research investigating the underlying mechanisms with which connected firms avoid tax. The preceding discussion suggests that the extent to which revenue authorities can constrain tax avoidance, as observed by Hoopes et al. (2012), could be subject to political influence and that connected firms are more tax aggressive, as observed by Kim and Zhang (2016), presumably because they can influence the probability of tax audits and the magnitude of post-audit adjustments and sanctions. To summarize, we expect the political influence of firms to mitigate the positive effect of tax authority enforcement in deterring tax avoidance. We state our composite hypothesis in the alternative as follows: HYPOTHESIS. Corporate political ties decrease the impact of tax authorities in constraining tax avoidance. Although our hypothesis is directional, we recognize there are countervailing incentives pushing politically connected firms to forgo current tax savings to reap other benefits later (Bradshaw et al. 2016; Erickson et al. 2004; Mills et al. 2013). There is also evidence casting doubt on the role of political relations in tax reporting. For example, the tax collection effort in China is influenced more by municipal leaders’ career advancement incentives and regional budgets, and less by firm-level political motivations (Chen et al. 2015; Deng and Luo 2011; Liu et al. 2015). To the extent that firms exploit political ties for priorities other than minimizing tax scrutiny, such ties would be unrelated to enforcement or even enhance the enforcement effect on tax payments. Moreover, as the political relationship is a personal asset that is based on reputational capital, connected board members have incentives to ensure that their personal reputation and goodwill in relation to the government is not eroded due to corporate tax avoidance. Such reputation concerns could make board members This article is protected by copyright. All rights reserved. reluctant to hinder tax enforcement. Whether political ties decrease tax authorities’ role in constraining tax avoidance is thus an empirical issue. 4. Research methodology Data and sample selection We focus on A-share firms listed on the Shanghai and Shenzhen stock exchanges and start with a sample of 15,070 firm-year observations from 2003 to 2013. We begin in 2003, the year following the 2002 tax revenue sharing reform and the implementation of strengthened rules on income transfers, and end in 2013, the last year complete tax enforcement data are available. We obtain firm financial data from the China Stock Market and Accounting Research (CSMAR) and WIND databases. We manually collect background information about top executives and board members from corporate annual reports. To prevent outliers from unduly affecting our results, we winsorize the top and bottom one percentile of all scaled variables and exclude (i) 157 firm-years in the financial industry; (ii) 2,824 firm-years with negative income tax expense or negative pre-tax income; (iii) 611 firm-years with an ETR equal to or greater than 1; (iv) 269 observations with no financial information to calculate the ETR; and (v) 88 observations from the Tibet Autonomous Region because only state tax is collected in this particular region. Our final sample contains 11,121 firm-year observations. Table 1 describes the sample distribution by year and by industry. Our sample firms cross multiple industries, with 50.87 percent in manufacturing and 9.25 percent in the real estate industry. Observations are fairly evenly spread across years. Untabulated results show that (i) listed firms are geographically dispersed with a maximum number of 132 in Guangdong and Shanghai, and a minimum of 10 in Qinghai and Ningxia; (ii) the country’s lowest-tax province is Ningxia (15.6 percent) and the highest is Liaoning (27.7 percent); and (iii) agriculture and real estate are the lowest and highest taxed industries, respectively (12.5 percent vs. 29.8 percent). This article is protected by copyright. All rights reserved. Model specification and definition of variables To test our hypothesis, we estimate the following regression model: ETRit = β0 + β1Enforcementkt + β2Connectedit + β3Enforcementkt×Connectedit + βkXikt + ԑit, (1) where i, k, and t are firm, region, and year indicators, respectively. ETR is the one-year effective tax rate, defined as the ratio of annual income tax expense to annual pre-tax income before special items (Hoopes et al. 2012). We consider whether our main results are sensitive to other measures of tax avoidance, including the one-year and three-year cash ETR and three-year book ETR, adjusted or unadjusted for size and industry effects (Balakrishnan et al. 2014; Kim and Zhang 2016). Tax enforcement We use aggregate data from the China Tax Audits Yearbook (2003–2013) to construct our tax enforcement measures. The yearbook is published annually by the SAT and contains detailed tax enforcement data for each province and major city in China. Specifically, the data include the number of permanent employees, tax inspectors, and employees with a bachelor’s degree and/or a professional qualification (certified accountants, certified tax agents, and lawyers) and their Communist Party membership and age range; corporate taxpayers; corporate tax returns audited; audit departments; suspicious cases; cases prosecuted; and cases closed. It also includes the amount of regional tax revenue, tax deficiencies settled, overdue tax surcharges, and tax penalties, interest, and fines. Following Hoopes et al. (2012), we examine the contemporaneous relation between enforcement and avoidance assuming that managers have rational expectations of the actual audit rates from public statistics. However, we find our results are robust to using the one-year lagged enforcement. This article is protected by copyright. All rights reserved. We perform principal component analysis to identify and interpret common factors. This procedure reduces the dimensionality of the data by combining the individual aspects of tax enforcement into a parsimonious set of linear combinations that explain the shared variance among the characteristics. Our analysis produces three common factors, which we label Probability, Expertise, and Outcome, which we then combine into an aggregate measure called Enforcement. Appendix 1 describes our factor analysis in detail. Figure 1 shows which provinces fall into each quartile of Enforcement for the period 2003–2013. Firms in regions with the darkest shade are most likely to be audited, confront a well-trained and skilled audit workforce, and face the toughest fines for underreporting. [Figure 1 here] Political connection To measure political connections of the board, we collect data from corporate annual reports. We classify board members as politically affiliated if they serve or formerly served in a city-level government agency (e.g., taxation/finance bureaus) or are current or ex-members of the People’s Congress or the People’s Political Consultative Conference (Fan et al. 2007).10 We consider both the existence and the strength of connections. We code ConnectedChair as one for firm-years with a politically connected chairman and zero otherwise. ConnectedBoard equals one if the firm-year’s number of connected board members exceeds the industry-year median and zero otherwise. We measure the strength of political connections for each firm/year with (i) the number of connected board members (Connected#), defined as the natural logarithm of one plus the number of connected members 10 Although the central government has ultimate power over local government decisions, the local governments have the incentive to help rent-seeking firms. Kim and Zhang (2016) also use corporate campaign contributions and corporate lobbying expenditures to construct the data set. As there are no opposition parties in China, all political connections are with government officials from the Communist Party and Consultative Party. This article is protected by copyright. All rights reserved. in the year; and (ii) the percentage of connected members on board (Connected%).11 Using a continuous variable assumes a log-linear relation between the variables of interest and avoids imposing a subjective and arbitrary cutoff point, whereas using an indicator variable relaxes this assumption and allows one to observe whether the slope of a variable differs across the levels of another variable. To test our hypothesis, we interact the political connection variables with the enforcement variables. The coefficient of Enforcement captures the difference in ETR for unconnected firms facing different levels of tax enforcement, whereas the sum of the coefficients of Enforcement and its interaction with the connection variables captures that for connected firms. A significantly negative coefficient for the interaction term will indicate political ties attenuate enforcement. We include multiple control variables known to affect tax avoidance (e.g., Chen et al. 2010; Gupta and Mills 2002; Kim and Zhang 2016; Mills et al. 2013; Wu et al. 2007). We include the following firm-specific characteristics: profitability (ROA), the ratio of net income to total assets, and its standard deviation (Std. dev. of ROA); firm size (Size), the natural logarithm of the firm’s year-end total assets; asset liquidity (Liquidity), the current assets over current liabilities; financial leverage (Leverage), the total debt over total assets; plant, property, and equipment (PPE), the ratio of net PPE to total assets; growth (Growth), the market value scaled by net book value of assets; intangibles (Intangible), the ratio of intangible assets to total assets; inventory (Inventory), the ratio of inventory to total assets; cash holdings (Cash), the ratio of year-end cash holdings to lagged assets; and ownership structure (Ownership), the percentage of equity held by the government. We include discretionary accruals (Accruals), estimated from the modified cross-sectional Jones model, to control for the effect of financial reporting aggressiveness (Frank et al. 2009). 11 The percentage of political connected boards is the same for a board with one connected member out of five or for another board with two connected members out of 10. However, as one may argue that the latter board has more political clout than the former, we present both scenarios. This article is protected by copyright. All rights reserved. We also include three factors to control for differences in regional economic and institutional conditions and the effect of local leader turnover (Chen et al. 2015): GDP, the natural logarithm of a province’s gross domestic product; Institution, an index score that reflects a region’s institutional characteristics (Fan et al. 2011); and LeaderTurnover, an indicator variable that takes the value of one if the province where the firm is headquartered experiences governor or party secretary turnover, and zero otherwise. Finally, we include year and industry fixed effects to control for macroeconomic conditions and changes in tax regulations that differ across years and industries. Appendix 2 summarizes the regression variable specifications. 5. Empirical results Descriptive statistics on tax enforcement and political ties Table 2 describes our dependent and explanatory variables. Mean ETR is 23.3 percent and varies substantially, but not due to connected firms clustering in lower-tax industries or being smaller in size, nor by regions with the above-median enforcement level having more tax-favored industries or smaller firms. We also describe the raw components of enforcement. On average, 1.6 percent of corporate tax returns filed are selected for an audit (Ratio 1), 1.2 percent are prosecuted (Ratio 2), 17 tax officials are in charge of each listed firm (Ratio 3), 92.4 percent of the audit workforce have at least a bachelor’s degree (Ratio 4), 8 percent of the personnel are professionally qualified (Ratio 5 to Ratio 9), and 1.5 percent of a region’s tax revenue comes from deficiencies, penalties, interest, and surcharges (Ratio 8 and Ratio 9). A typical corporate board has about 15 directors, and about 66 percent have at least one connected director. The average number and percentage of connected board members are 1.28 and 8.75 percent. This article is protected by copyright. All rights reserved. Untabulated descriptive statistics suggest enforcement varies across region and time. For example, the highest audit rate is 15.7 percent for Hubei and 0.04 percent for Inner Mongolia. Jilin and Hainan are far more effective than other provinces in settling tax deficiencies, while Gansu and Ningxia impose heavier fines for tax noncompliance than Shanghai and Shandong. However, the proportion of the tax administration’s workforce with a university degree and/or a professional qualification does not differ significantly across regions. In addition, the three tax enforcement factors are highly correlated, suggesting that although each of the three factors captures different aspects of tax enforcement, they classify the provinces in a similar way. We match regional enforcement with the location of a listed firm’s headquarters, because this is where many key business decisions are made and is thus often the focus of inquiry by tax authorities. Further, the local tax administration in the headquarters’ region can change the tax owed by subsidiaries in other regions, so the strength of the tax office where a firm is headquartered matters. Our headquarters are widely dispersed across the six regions (the Southeast, Bohai, Central, Northeast, Southwest, and Northwest). Substantial cross-office variations in enforcement intensity ensure that our statistical tests have sufficient power to assess the effect of government scrutiny on corporate tax reporting. Univariate tests To provide univariate tests of our hypothesis, we rank all provinces in ascending order of Enforcement, group them into quartiles, and partition the sample at the median of Connected#. We compare the mean ETR of the two groups of firms within each of the four quartiles. Panel A of Table 3 shows that in areas with the lowest enforcement level (1st quartile), political connections do not explain variation in effective taxes. However, well-connected firms in the highest enforcement regions (3rd and 4th quartiles) report significantly lower mean ETR, consistent with our prediction. This article is protected by copyright. All rights reserved. Panel B shows that both connected and unconnected firms are dispersed across regions with different levels of tax enforcement. Panel C shows the descriptive statistics and univariate test results for control variables. For the full sample, connected firms report significantly larger return on equity (ROA), total assets (Size), inventory (Inventory), cash (Cash), and discretionary accruals (Accruals), but significantly smaller amounts of property, plant, and equipment (PPE), lower growth opportunities (Growth), and government-owned shares (Ownership).12 The differences in firm attributes between the two groups suggest the need for multivariate analysis. Multivariate tests We include the three enforcement variables (Probability, Expertise, and Outcome) separately (columns 1–3 of Table 4) and together (column 4), and then use our aggregate measure Enforcement (column 5).13 Across all specifications, the enforcement variables are positive and statistically significant. To determine which enforcement measure has relatively more explanatory power, we perform Vuong (1989) tests by running the baseline without any enforcement variables and then comparing its R2 value to the ones with each of the three variables. Untabulated results indicate that the consequence of tax audits has the 12 There are two opposing directional predictions about the tax sensitivity of SOEs (Cui 2015). The first is that because taxing SOEs simply involves the transfer of money from one pocket of the government to another, SOEs are insensitive to income tax and hence do not pursue tax avoidance. The second prediction is that SOE managers, like managers of private firms, are tax-averse; therefore, SOEs (even those that are wholly owned) are likely to be tax-sensitive and do engage in tax avoidance to increase after-tax cash flow. If we expect businesses to be sensitive to taxation, then excluding tax-insensitive (tax-sensitive) SOEs from our sample will enhance (weaken) our results. Nonetheless, our main results are invariant to the exclusion of SOEs from our sample. The central government’s share of corporate tax revenues collected from local SOEs increased from 0 percent before 2002 to 50 percent in 2002 and 60 percent thereafter, which may have induced local SOEs to shift income to lower-tax subsidiaries to increase after-tax cash flows. However, Shevlin et al. (2012) find no evidence that income shifting varies as a function of whether the largest shareholder is either the central or local government, either before or after 2002. Our sample period includes no policy changes in the tax-sharing rule. 13 As the same firm can appear several times in our sample and the residuals may be correlated across observations, we use the Huber-White standard errors clustered at the firm level for all regressions (Petersen 2009). For simplicity, we do not report the yearly and industry indicator variables. The highest Pearson correlation among the independent variables is −0.508 between PPE and Inventory and 0.533 between GDP and Institution. The remaining correlations are all below 0.40, which is far below 0.80, the point beyond which the threat of multicollinearity becomes a real concern (Judge et al. 1988). This article is protected by copyright. All rights reserved. most incremental ability to explain variation in ETR, followed by the probability of audits, although the differences are small. This is reasonable because the decision of whether to initiate an audit and to impose post-audit penalties (if so, how much) is more vulnerable to external influences.14 The weaker results on auditor expertise could be due to insufficient year-to-year variation in auditors’ personal characteristics (e.g., academic degree and professional qualifications) or more expert auditors being less vulnerable to influence. Column 4 shows all three variables load similarly to the separate models (columns 1–3), in spite of collinearity. Our coefficient estimate on Enforcement in column 5 implies that raising the enforcement intensity from 37 percent (the 25th percentile in our data) to 61 percent (the 75th percentile) increases ETR, on average, by 1.1 percent, translating into an extra annual tax payment of USD 1.08 million (equivalent to 5.08 percent of the reported tax expense) for the average firm. Therefore, we conclude that Chinese tax authorities substantially deter tax avoidance, consistent with the results in Hoopes et al. (2012) for U.S. data. The sign and significance of the control variable coefficients are generally consistent with prior literature (Chen et al. 2010; Hanlon and Heitzman 2010; Kim and Zhang 2013; Mills et al. 2013; Wu et al. 2007). Specifically, higher ETRs are associated with firms reporting higher levels of cash, inventory, and tangible and intangible assets, and locating in regions with more developed economies and less mature institutional environments, but reporting lower levels of net income and discretionary accruals. 14 Because our audit data are regional, not firm-specific, we cannot examine whether connected firms are associated with a lower probability of being audited, although anecdotal evidence suggests so (Shevlin et al. 2012). To shed light on whether connected firms face lesser penalties conditional on an enforcement action being taken, we hand-collect government penalty data from financial statement footnotes. We identify 6,460 firm-years with misconduct penalties ranging from tax avoidance and evasion, environmental pollution, food safety, product defects, security fraud, and false information disclosures. In untabulated results, we find firms with a connected board enjoy smaller penalties (scaled by assets). This article is protected by copyright. All rights reserved. To test our hypothesis, we interact the three enforcement measures with the political connection variables. However, for brevity, in Table 5 and throughout the rest of the paper, we present the results using the aggregate Enforcement variable, rather than the separate enforcement measures. 15 Columns 1, 2, 3, and 4, respectively, use ConnectedChair, ConnectedBoard, Connected#, and Connected% for political connectedness. Across columns 1–4, Enforcement explains higher ETR, consistent with Table 4. Consistent with our hypothesis, the coefficient on Enforcement×Connected is negative and significant, which implies a weakening of the responsiveness from 0.059 to 0.066 for unconnected firms to about 0.008 for connected firms, a drop of 86 percent to 88 percent (see columns 1–2 and the F-statistics therein). The coefficient on the interaction variable in column 4 implies that a one standard deviation increase in the percentage of connected board members is associated with a decrease in ETR of 0.059 (= –0.002×9.244 percent). This value is larger in magnitude than the coefficient on Enforcement in the last column of Table 4 (i.e., 0.047), which suggests that such an increase in political ties would entirely offset the positive effect of tax enforcement on tax costs. Collectively, these results are consistent with our prediction that connected firms can make the tax authority more lenient in enforcing corporate tax reporting or that the revenue agency experiences greater difficulty in enforcing tax compliance for firms with a connected board. Results for the control variables in Table 5 are generally consistent with those in Table 4. 15 As a zero value of enforcement is not meaningful, to increase interpretability we center Enforcement first (i.e., subtract the annual mean from each case) and then compute Enforcement×Connected. Centering does not change what a model means or what it predicts, but by centering we can easily interpret the effect of political connections for firms in regions with an average level of enforcement. Further, if the variable is not centered, its product used in computing the interaction may be highly correlated with other variables, thereby inducing multicollinearity (Aiken and West 1991, 32–33). For example, the correlation between Enforcement×ConnectedBoard and ConnectedBoard is 0.941 before centering and 0.007 after centering. The average variance inflation factors (VIFs) also reduce from 3.49–3.57 to 2.95–2.98 in our centered models. This article is protected by copyright. All rights reserved. Changes in political connections To strengthen our analysis and help address the endogeneity concern related to omitted variables (e.g., corporate attitudes toward politics and taxation and government spending on tax education), we employ a DID procedure to analyze the effect of a political connection status change. Such a change could arise from either unknown firm-specific preferences changing, or from an exogenous shock to corporate political ties. For brevity, we do not report the results for control variables. First, we examine the effects of new politically connected director appointments. We identify 242 firms that change their connection status from zero (i.e., ConnectedBoard = 0) to one (i.e., ConnectedBoard = 1) and 595 firms that remain politically unconnected during our sample period. We match each treatment firm (i.e., with a connection status change) with an unconnected control firm based on firm size (using the logarithm of sales revenue to identify the closest firm) and ownership (SOEs and non-SOEs).16 After matching, we have 1,828 × 2 firm-years. We then estimate the following DID regression: ETRit = β0 + β1Enforcementkt + β2Treatmenti + β3Postt + β4Enforcementkt×Treatmenti + β5Enforcementkt×Postt + β6Treatmenti×Postt + β7Enforcementkt×Postt×Treatmenti + βkXikt + ԑit, (2) where Treatment is an indicator variable that equals one for firms that increase their political connections, and zero otherwise. Post equals one for firm-years in the post-change period, and zero otherwise. Using this specification, we can investigate whether tax enforcement becomes less effective for firms that newly obtain political connections than for control firms that remain politically unconnected. Thus, by using a matched control sample, we control for the multiple other factors not related to the connection change that could be responsible for We considered allowing 1:N matching to increase our power and gain more confidence in our inferences. However, due to a small pool of control firms we are unable to expand the 1:1 to even a 1:2 match for our treatment firms, so we only implement this test with a 1:1 treatment-control group. 16 This article is protected by copyright. All rights reserved. the difference in the effect of enforcement on ETR. Our interpretation of the primary coefficients follows. The coefficient on Enforcement (β1) measures whether government enforcement affects ETR. A positive Enforcement coefficient suggests higher tax enforcement is associated with higher effective taxes for control firm-years in the pre-change period. The interaction term Enforcement×Treatment (β4) captures any cross-sectional variation in enforcement effectiveness between treatment and control firms pre-change. Enforcement×Post (β5) captures the temporal trend in enforcement effectiveness for control firms, unrelated to a change in the board’s connection status. As the board’s connection status is the same for both groups of firms pre-change or remains unchanged for control firms post-change (meeting the parallel trends assumption), we expect β4 and β5 to be insignificant. 17 Finally, Enforcement×Treatment×Post (β7) captures how the estimated coefficients on Enforcement evolve over time for treatment firms (i.e., β5 + β7) or for treatment versus control firms (i.e., β4 + β7). A negative β7 indicates firms that obtain political connections (i.e., treatment firms) enjoy reduced enforcement effectiveness compared to those that remain politically unconnected (control firms). Table 6, panel A1, reports the DID regression results. Panel A2 presents a two-by-two matrix to illustrate changes in enforcement effectiveness over time for control versus treatment firms.18 We conduct an F-test to determine whether a group of variables is jointly 17 To better meet the parallel trends assumption, we also use a propensity score matching approach. We estimate a probit model using Treatment as the dependent variable and several fundamental firm characteristics (Size, ROA, Leverage, Liquidity, Growth, and SOEs) and year fixed effects as the explanatory variables. We use the predicted propensity score to match each treated observation with the control observation. This procedure generates 278 matched pairs. The t-test results for the between-group differences show that the two groups of firms are very similar in fundamental firm characteristics, indicating that our matching is effective. We re-estimate this model and find that Enforcement has a significantly positive coefficient (0.152, t = 2.14), and more importantly, our DID (Enforcement×Treatment×Post) coefficient continues to be negative and significant (−0.645, t = −1.84). 18 We thank a reviewer for suggesting this matrix to help readers interpret our results. Although the DID method produces estimates of changes that are more plausible than those based on a single difference, results from triple interactions are harder to interpret. To simplify the analysis, we regress ETR on Enforcement, Post, Enforcement×Post, and control variables based only on the sample of firms that changed their political connection status from zero to one. We find (untabulated) that the coefficients are positive for Enforcement (t-stat = 1.93, p-value = 0.053) and negative for the interaction variable (t-stat = −2.41, p-value = 0.016), consistent with the DID results. This article is protected by copyright. All rights reserved. significant. From the second row/column of the matrix, we observe that enforcement effectiveness declines for firms that evolve from having no political connection to acquiring a new connection.19 This suggests that firms gain tax-favored treatment in the enforcement process following the establishment of a political connection. Second, we examine whether a plausibly exogenous change in the political landscape moderates the sensitivity of ETRs to a tax audit. Corruption has plagued China since the economic reforms started in the late 1970s. To crack down on “tigers” (high-level officials) and “flies” (local civil servants), President Xi Jinping initiated an unprecedented countrywide anti-corruption campaign that has prosecuted many corrupt officials since 2012. 20 The campaign signaled an exogenous shock to the rent-seeking and favor-exchanging ability of both politicians and managers. Undoubtedly, many firms lost their established political ties as a result of the post-2011 crackdown on corruption. To examine the effect of this political environment change, we estimate the following regression: ETRit = β0 + β1Enforcementkt+β2ConnectedBoardit + β3Post2011t + β4Enforcementkt×ConnectedBoardit + β5Enforcementkt×Post2011t + β6ConnectedBoardit×Post2011t + β7Enforcementkt×ConnectedBoardit×Post2011t + βkXikt + ԑit, (3) 19 We also consider comparing the difference between firms that undergo a change in their connection status from one to zero and firms that remain politically connected. We find 175 firms that experience such a change during our sample period. However, matching unconnected firms with their counterparts on the basis of firm size and ownership significantly reduces our sample size. We fail to find an incremental increase in the Enforcement coefficient following political disconnections, probably because the disconnection is due to the regular turnover (e.g., retirement) of board members, which does not immediately terminate the existing bureaucratic ties. This is why we next turn to a more powerful test of the effect of political cessations caused by extrinsic events. 20 According to the Corruption Perceptions Index reported by Transparency International, the corruption index in China averaged 33.7 points from 1995 to 2015, reaching an all-time high of 40 points in 2013. The Chinese Communist Party’s 18th Congress held in November 2012 marked a key power transition. The official records reveal that more than 336,000 party officials at various levels have been investigated or arrested, including 98 vice ministers or officials above that level (http://www.transparency.org/whatwedo/publication/cpi_2014). This article is protected by copyright. All rights reserved. where Post2011 equals one for the 2012–2013 observations and zero for the 2010–2011 observations.21 The incremental cross-sectional and time-series variations in the Enforcement coefficients are captured by β4, β5, and β7. β4 indicates the incremental difference in the effect of enforcement on ETR between connected and unconnected firms pre-2011. A negative sign on β4 is consistent with political ties influencing tax audit outcomes more favorably for connected firms. β5 captures any temporal change in the effect of enforcement for firms that are not influenced by regime change. We have no prediction on the sign of this coefficient. β7 captures the incremental difference in the enforcement effectiveness between connected and unconnected firms in the post- versus pre-2011 period (i.e., β4 + β7 vs. β4) or between the two periods for connected versus unconnected firms (i.e., β5 + β7 vs. β5). Our predicted sign for this DID coefficient (β7) is positive, consistent with political regime change causing connected firms to lose some of their ability to influence government decisions in their favor. Table 6, panel B1, reports the DID regression results, and panel B2 presents a two-by-two matrix to show the sensitivity of tax to enforcement classified by political connectedness in both pre- and post-2011 periods. While environmental change has little effect on enforcement effectiveness for unconnected firms (first row), its effect on connected firms is substantial (second row). The two columns of the matrix convey a similar message: following a regime change in 2012, the political unwinding of the deterrent effect of tax enforcement has completely diminished (β4 + β7 = 0.046, F-stat = 0.60, p-value = 0.438). Collectively, our cross-sectional and cross-temporal comparisons reinforce our main results that the impact of government agencies in constraining tax avoidance weakens after a firm’s political connection is established and strengthens after this connection is severed/scrutinized. 21 The decision regarding the time frame to examine the effect of an event involves a trade-off: using a longer window will include more observations thereby increasing the statistical power but will also increase the confounding effect of other uncontrolled factors (as the environment changes with time), while using a shorter window will minimize the confounding effect but will include fewer firm-years. We balance this trade-off by comparing two years before and two years after the event. Our coefficients of interest remain substantially unchanged, both in levels and in statistical significance, if we use the full sample. This article is protected by copyright. All rights reserved. These results also provide some assurance that our core results are unlikely to suffer from an endogeneity bias. Self-selection bias Our results so far show that the strong positive association between ETR and tax enforcement is moderated by political connectedness. A serious concern, however, is that our results may be biased if a firm’s decision to become connected is affected by its tax burden.22 As firms choose whether and how deeply to connect with politicians, it is possible that a comparison of the tax sensitivity for firms with versus without a connection biases the estimated sensitivity. To alleviate the concern of the endogenous choice of establishing political ties, we employ the Heckman two-stage model, similar to Kim and Zhang (2016). Lennox et al. (2012) stress the importance of imposing exclusion restrictions in implementing the Heckman model. To justify the validity of the exclusion restriction, we follow Kim and Zhang (2016) by using an instrumental variable, Industry % of Connected Firms, defined as the percentage of firms with a politically connected board in a firm’s industry group. In the first stage, we regress ConnectedBoard against Industry % of Connected Firms and the same set of control variables from Table 4 using probit analysis. Table 7 shows that the instrumental variable loads significantly positive (coeff. = 3.043, t-stat = 7.82). In spite of our instrument being strongly significant, we acknowledge that the overall explanatory power of the first stage is modest (R2 = 6 percent).23 Other results 22 We think that a firm’s decision to select a headquarters location is less likely to be affected by its tax burden. An individual firm is less likely to be able to influence provincial-level tax enforcement through its own effort, so we do not expect enforcement to be endogenous in our analysis. Tax enforcement would be endogenous to firm-specific avoidance only to the extent that one single firm is able to influence tax audit decisions in the region (see Hoopes et al. 2012 for a similar argument), or firms respond to enforcement intensity by relocating their headquarters from region to region. However, corporate location decisions are more likely to be driven by exogenous factors (e.g., proximity to customers, suppliers, and production inputs) than by decisions to escape tax investigation (Hoopes et al. 2012; Loughran 2008); and once determined, the location of the corporate headquarters seldom changes. 23 To improve the R2 value, we consider other predictors known to be associated with political ties but not with effective taxes. We employ distance to the provincial capital because the closer the capital, the greater the likelihood to enter politics (e.g., Guedhami et al. 2014; Kim and Zhang 2016). We consider market development and business competition at the regional level because as markets develop, competition between This article is protected by copyright. All rights reserved. indicate that larger and more profitable firms and firms with less leverage and a lower government stake are more likely to seek political ties. In the second stage, when we include the inverse Mills ratio (Inverse Mills Ratio), the coefficient on Enforcement×ConnectedBoard continues to load negatively at the 5 percent level (coeff. = −0.058, t-stat = −2.18). Notwithstanding the foregoing analysis, we stress that we do not attempt to identify a causal effect of political influence on tax burdens. Rather, we test whether such influence affects tax avoidance by shaping the enforcement action of tax authorities, and we conduct our test through an interaction variable. We think the interaction term is less prone to induce endogeneity, unless the individual firm has full control over the joint outcome of the two variables. The mechanism through which firms avoid taxes We next focus on the most common type of tax aggressiveness within China—domestic income shifting within a book-consolidated business group—to examine whether political relations weaken the effectiveness of enforcement against tax-motivated income shifting.24 Separate tax reporting by member firms creates strong tax incentives for them to shift income firms intensifies and the uncertainty facing firms increases, motivating firms to develop ties to bureaucrats in order to reduce this uncertainty (Haveman et al. 2017). Regrettably, we do not find significant improvement to the R2 value. As the interpretations of the significant variables are the same for both high and low R2 models, we instead pay more attention to the appropriateness of the IV and the presence of multicollinearity in the second-stage model. Our IV is an industry-level variable exogenous to an individual firm’s influence; its effect on the first-stage dependent variable is significant (coeff = 3.043, t = 7.82); and it has low correlation with the ETR (0.009), suggesting this IV is valid according Lennox et al. (2012) and Larcker and Rusticus (2010). We also find that the VIFs are 1.08 for ConnectedBoard and 9.07 for Inverse Mills Ratio, implying that the model does not face serious multicollinearity problem according to Greene (2008). Nonetheless, we acknowledge our compromises and limitations in pursuit of answering interesting questions (Bloomfield et al. 2016) about the effect of political ties. 24 Most income-shifting studies have been done at the international level. Subnational research has the advantage that, while tax rates may differ across regions and industries, nontax factors are more homogenous between regions than between nations (Gupta and Mills 2002). Relative to multinational income shifting, subnational shifting is less costly and easier to execute because income shifters are not constrained by import and export tariffs, foreign exchange controls, or dividend repatriation restrictions. Although prior studies provide extensive evidence on the tax effect of both types of income shifting, our study permits granular detail on enforcement and political connections that would be much harder to obtain across multiple countries. This article is protected by copyright. All rights reserved. within the group. Chinese tax authorities employ many ways to detect and deter tax avoidance via transfer pricing. For example, recent tax returns require nine transfer pricing-related forms that report the amounts and types of related party transactions and the methodologies the firm uses to arrive at an arm’s length price. When selecting targets for transfer pricing audits, tax authorities focus on firms with the following characteristics: (i) a significant amount or numerous types of related party transactions; (ii) consecutive losses, low profitability, or a fluctuating pattern of profit/loss; (iii) profitability lower than others in the same industry; (iv) business dealings with a related party in a tax haven; and (v) lack of tax return disclosures relating to transfer pricing. 25 However, they pay less attention to domestic income shifts that are taxed at the same rate (SAT , Circular 2). Tax audits conclude with a final assessment that includes the proposed adjustment, interest levy, and penalties (Chan et al. 2010). The high ex ante threat of a tax audit, heavy penalties, and ex post incidence of closer monitoring should deter firms from taking aggressive tax positions. We obtain domestic related party operating transactions from the CSMAR database. Using the firm-years for our main results as the starting point, we delete observations with no related transactions and no tax rate differentials within the group. We also exclude observations whose related party transactions exceed consolidated total assets. Our reduced sample for this test contains 7,869 observations. Following Jacob (1996), we model a firm’s opportunity to shift income for tax reasons as a joint function of the spread in tax rates among members of the consolidated group and the magnitude of intragroup transactions. This design is important because firms would have little tax incentives to shift profits if such rate differences do 25 In the United States, the IRS instructs its agents to scrutinize firms reporting significant book-tax differences (U.S. Department of the Treasury 1999). Mills (1998) documents a positive association between book-tax differences and proposed tax audit adjustments. The largest U.S. firms are subject to a higher audit rate (Hoopes et al. 2012). In the Chinese context, Lennox et al. (2015) find that firms are more likely to receive a tax audit if they report low ETRs, large book-tax differences, and negative discretionary accruals. Chan et al. (2010) find that the larger the difference between book and tax incomes, the larger the adjustment of tax audits. Lin et al. (2014) find that private firms are more likely than public firms to make income-decreasing accruals when the tax rate is about to decrease. To the extent that income-decreasing accruals signal a potential underreporting of taxable incomes, they are more likely to trigger a tax audit. However, our book-based ETR measure will not directly address avoidance via book-tax differences. This article is protected by copyright. All rights reserved. not exist; or even if they do, group firms have little or no intragroup transactions (we find neither component alone is sufficient to explain ETRs). We estimate the following model to examine whether enforcement constrains tax-induced income shifting and political influence reduces that constraint: ETRit = β0 + β1Shiftingit + β2Enforcementkt + β3ConnectedBoardit + β4Shiftingit×Enforcementkt + β5Shiftingit×ConnectedBoardit + β6Enforcementkt×ConnectedBoardit + β7Shiftingit×Enforcementkt×ConnectedBoardit + βkXikt + ԑit, (4) where Shifting is a product of the difference between the highest and lowest statutory tax rates faced by any group member (i.e., the rate range) and the aggregate dollar amount of related transactions (scaled by the lagged assets of the group). As the rate range is highly skewed with a break in the distribution around 90th percentile, prior to computing Shifting we transform the range into a dummy variable that equals one if the group is in the top decile of the rate range and zero otherwise. To facilitate interpretation, we convert Enforcement into an indicator variable: for each year during our sample, we classify provinces with the enforcement value falling into the upper quartile of the sample distribution as regions with strong tax enforcement (= 1) and the remaining three quartiles as weak enforcement regions (= 0). We obtain (untabulated) the same inferences when we use a continuous variable. Other variables are as previously defined. Our ex ante prediction about the coefficients of interest is as follows. The baseline coefficient measures the association between Shifting and ETR for unconnected firms in weak enforcement regions; a negative β1 means greater Shifting is associated with lower ETR. β4 measures the incremental effect of tax enforcement and should have a positive sign if unconnected firms with greater Shifting are associated with incrementally smaller declines in ETR when enforcement becomes stronger. β5 indicates a change in this association for connected firms versus unconnected firms in regions with weak enforcement. We have no prediction on the sign of this coefficient because firms have less of This article is protected by copyright. All rights reserved. a need to exploit their political ties for tax benefits when government enforcement is ineffective. The main coefficient of interest for testing whether political ties weaken the effectiveness of government enforcement against tax-induced income shifting is β7. This represents the incremental association between Shifting and ETR for connected versus unconnected firms in strong enforcement regions or for connected firms in regions with strong versus weak enforcement. Our predicted sign for this coefficient is negative, consistent with firms establishing political ties to influence the outcomes of tax enforcement actions in their favor. Table 8 (panel A) reports how tax enforcement and political connections separately and jointly moderate the firm’s incentive to shift income for tax purposes. Across all specifications, the coefficients that exhibit statistical significance in the predicted direction are β1, β2, β4, and β7. They indicate that government enforcement mitigates or inhibits income-shifters’ capability to avoid tax, while political ties enhance it. Panel B presents whether the Shifting coefficients vary according to political connectedness and enforcement strength. The matrix shows evidence similar to our core results: corporate political ties can make tax enforcement less effective in constraining tax avoidance (an incremental difference of 0.018), or put differently, government enforcement is less effective for firms with a politically connected board (an incremental difference of 0.160). 6. Sensitivity tests and other supplemental analysis We evaluate the sensitivity of our main results to different definitions of our dependent and test variables. For brevity, we do not tabulate these results. This article is protected by copyright. All rights reserved. We first confirm that our results are not sensitive to alternative measures of ETR for the dependent variable. We estimate annual cash ETRs (CashETR) using the ratio of cash taxes paid to pre-tax income.26 We estimate the three-year book ETR (3-year BookETR) using the ratio of the three-year sum of income tax expense to the three-year sum of pre-tax income in years t, t+1, and t+2. Using 3-year BookETR avoids significant year-to-year variation in the annual ETR (Dyreng et al. 2008; Kim and Zhang 2016) and also reflects the likelihood that the tax bureau may audit additional years if it discovers underreporting in a certain year. We similarly estimate a 3-year CashETR. We also use 3-year adjusted BookETR following Balakrishnan et al. (2014) and Kim and Zhang (2016) by subtracting each firm’s 3-year BookETR from the average 3-year BookETR for firms in the same industry and quartile of total assets. An unusually low ETR compared to similar firms can be viewed as more tax aggressive (Balakrishnan et al. 2014), and a positive value of the adjusted ETR implies more tax avoidance because that firm reports less tax than its size-industry peers. Our results are not sensitive to these alternative ETR measures. We also evaluate alternative measures of the explanatory variables. Our results are robust to using the one-year lagged enforcement data or to using the raw enforcement data rather than the more complex principal component analysis. Results are qualitatively similar when redefining Enforcement using the number of income tax audits conducted scaled by the number of income tax returns filed in a given region/year, the number of income tax cases prosecuted scaled by the number of income tax returns filed in a given region/year, the percentage of tax officers with a professional qualification, or the sum of additional tax payments and penalties, interest, and surcharges scaled by regional tax revenues. 26 As Chinese firms do not disclose cash income tax payment information, we estimate this amount by taking total tax expense plus beginning taxes payable minus ending taxes payable, following Bradshaw et al. (2016). This article is protected by copyright. All rights reserved. Finally, we apply propensity score matching to help dispel the concern that differences in firm attributes between firms with and without political relations are spuriously responsible for our core results. We first estimate a logistic regression using ConnectedBoard as the dependent variable.27 Using a predicted propensity score from this logistic regression, we then match (without replacement) each connected firm with a nonconnected firm based on the closest propensity score within a caliper distance of 10 percent. We are able to identify 3,528 matched pairs. Our main results are invariant to this control for endogeneity issues associated with political relations. 7. Conclusion There has been a call to contribute to the scant literature examining the interaction between politics and tax enforcement agencies susceptible to political influence. We answer this call directly by studying whether political ties in the form of board interlocks weaken the ability of revenue agencies to hamper tax avoidance. Using aggregate government data, we develop a new composite measure of tax enforcement that captures the geographic and annual variation in the probability of detecting and prosecuting tax avoiders, the sufficiency and expertise of tax auditors, and the consequences and penalties of tax noncompliance. We find that while audit coverage and penalty rates deter an individual firm from tax avoidance, their deterrent effects are significantly undermined in the presence of a politically connected board of directors. Using an event study approach that mitigates the concerns over endogeneity, we find robust evidence that the ability of firms to affect their tax burden is higher following the addition of politically affiliated board members, but lower after a change in political environment that constrains the ability of politicians and firm managers to exchange favors. The independent variables include various firm characteristics (i.e., Industry % of Connected Firms, cross-listing, size, profitability, leverage, government shareholding) known to affect the likelihood of political ties in the prior literatures (e.g., Faccio 2006; Kim and Zhang 2016). We also expand our model to include other firm characteristics (e.g., growth, accruals, cash holding) and variables that proxy for regional economic and institutional conditions. In this extended model, we match 3,516 pairs and obtain very similar results (untabulated). 27 This article is protected by copyright. All rights reserved. Our study contributes to the scarce literature by providing evidence that because of political influence, revenue agencies subject the tax aggressiveness of connected firms to less scrutiny and light punishment, making their tax burden significantly lower than the statutory rate. To better understand the mechanism underlying tax avoidance, we use income-shifting opportunities to shed light on whether tax enforcement constrains this form of noncompliance while political connections assist it. We exploit the unique situation in which differently taxed entities within a consolidated group file their tax returns as independent legal entities in China to examine the extent to which taxation motivates the shifting of income within the affiliated group. We find evidence consistent with our main hypothesis. This additional analysis contributes to the very limited amount of literature addressing the combined effects of tax-induced income shifting, the deterrent role of tax authorities in this shifting, and the political unwinding of this role. Politically controlled economies represent a significant portion of the world economy (Faccio 2006), and the role of politics in shaping the government-business relationship is much more salient in the relations-based Chinese environment than the market-based U.S. environment. From a policy viewpoint, our findings imply that because the administrative effectiveness of tax systems in developing countries to protect revenues is influenced by both formal (e.g., the justice system and tax enforcement) and informal (e.g., personal relationships and social networks) institutions, successful tax reform should not be limited to strengthening the tax administration within the revenue system. Reforms must also encompass a country’s political and business environments, including government-business relations, the political economy of regulatory agencies, and the incentives and behavior of different actors within and outside the tax system. Our study faces several limitations, each of which represents a potential research agenda. First, as our analysis is based on the conventional estimation of ETR, the usual caution This article is protected by copyright. All rights reserved. applies in interpreting the results (Hanlon and Heitzman 2010). For example, our estimation of ETR does not pick up conforming tax avoidance because the measure uses book income as the denominator and we cannot generalize our inferences to unprofitable firms because firms with negative pre-tax income are excluded. Second, because the tax authority does not publicly disclose firm-level audit data, our tests of political connections on tax enforcement are based on aggregated audit data and hence are indirect. Third, to the extent that our empirical strategies are unlikely to completely solve the endogeneity issue, we urge readers to be cautious in interpreting our results. Finally, although we try to ensure that political connections are sought by the firm, rather than imposed by the government as in SOEs, we are unable to tease out the effect of government ownership from that of political connections. As most Chinese listed firms have some government ownership, local governments play dual but conflicting roles as both state tax collector and residual income claimant of their controlling firm. This represents a study in itself: how does the government enforce taxes when it is also the claimant of a firm’s before- and after-tax income? Future work can examine how local governments in a transition economy such as that of China trade off their tax collection efforts and tax avoidance incentives. This article is protected by copyright. All rights reserved. References Adhikari, A., C. Derashid, and H. Zhang. 2006. Public policies, political economies, and effective tax rates: Longitudinal evidence from Malaysia. Journal of Accounting and Public Policy 25: 574–595. Aiken, L. S., and S. G. West. 1991. Multiple regression: Testing and interpreting interactions. SAGE Publications, London. Altshuler, R., and H. Grubert. 2001. Where will they go if we go territorial? Dividend exemption and the location decisions of U.S. multinational corporations. National Tax Journal 54 (4): 787–809. Atwood, T. J., M. S. Drake, J. N. Myers, and L. A. Myers. 2012. Home country tax system characteristics and corporate tax avoidance: International evidence. The Accounting Review 87 (6): 1831–1860. Bagchi, S. 2012. Are Republican Administrations light on tax enforcement? A look at the IRS. Working paper, University of Michigan. Balakrishnan, K., J. Blouin, and W. Guay. 2014. Does tax aggressiveness reduce financial reporting transparency? Working paper at SSRN: http://dx.doi.org/10.2139/ssrn.1792783. Benari, G. 2016. The political economy of tax enforcement: a look at the Internal Revenue Service from 1978 to 2010. Journal of Public Policy 36 (3): 335-380. Bloomfield, R., M. W. Nelson, and E. Soltes. 2016. Gathering data for archival, field, survey, and experimental accounting research. Journal of Accounting Research 54 (2): 341-395. Bradshaw, M., G. Liao, and M. Ma. 2016. Ownership structure and tax avoidance: Evidence from agency costs of state ownership in China. Working paper at SSRN: http://ssrn.com/abstract=2239837. Brown, J. L., K. Drake, and L. Wellman. 2015. The benefits of a relational approach to corporate political activity: Evidence from political contributions to tax policymakers. Journal of the American Taxation Association 37 (1): 69-102. Bushee, B.J. 1998. The Influence of Institutional Investors on Myopic R&D Investment Behavior. The Accounting Review 73 (3): 305-333. Cai, H., and Q. Liu. 2009. Competition and corporate tax avoidance: Evidence from Chinese industrial firms. The Economic Journal 119 (April), 764–795. Chan, K. H., K. Z. Lin, and P. L. Mo. 2010. Will a departure from tax-based accounting encourage tax noncompliance? Archival evidence from a transition economy. Journal of Accounting and Economics 50: 58–73. Chaney, P. K., M. Faccio, and D. C. Parsley. 2011. The quality of accounting information in politically connected firms. Journal of Accounting and Economics 51: 58–76. Chen, S. X. 2014. Political determinants of selective tax enforcement: Evidence from China. London School of Economics and Political Science Working Paper. Chen, S., X. Chen, Q. Cheng, and T. Shevlin. 2010. Are family firms more tax aggressive than non-family firms? Journal of Financial Economics (95): 41–61. Chen, Z., Dyreng, S. D., Li, B. 2015. Corporate political contributions and tax avoidance. Working paper at SSRN: http://ssrn.com/abstract=2403918. Chen, D., S. Li, J. Z. Xiao, and H. Zou. 2014. The effect of government quality on corporate cash holdings. Journal of Corporate Finance 27: 384-400. Chen, H., S. Tang, D. Wu, and D. Yang. 2015. The political dynamics of corporate tax avoidance: The Chinese experience. Working paper at SSRN: http://ssrn.com/abstract=2640111. Chen, S., Y. Wang, and Z. Zhao. 2009. Regulatory incentives for earnings management through asset impairment reversals in China. Journal of Accounting, Auditing & Finance 24: 589–620. This article is protected by copyright. All rights reserved. Correia, M., 2014. Political Connections and SEC Enforcement. Journal of Accounting and Economics 57: 241-262. Cui, W. 2015. Taxation of state-owned enterprises: A review of empirical evidence from China, in Benjamin Liebman and Curtis Milhaupt (Eds.), Chinese State Capitalism and Institutional Change: Domestic and Global Implications, Oxford University Press. Crocker, K., and J. Slemrod. 2005. Corporate tax evasion with agency costs. Journal of Public Economics 89: 1593–1610. Dechow, P.M., R.G. Sloan, and A.P. Sweeney. 1996. Causes and Consequences of Earnings Manipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC. Contemporary Accounting Research 13 (1): 1-36. Deng, Y. H., and T. Luo. 2011. Tax revenue manipulation by local taxation administration in China. Asia-Pacific Journal of Accounting and Economics 18: 61–76. Desai, M. A., A. Dyck, and L. Zingales. 2007. Theft and taxes. Journal of Financial Economics 84 (3): 591–623. Dharmapala, D., and J. Hines. 2009. Which countries become tax havens? Journal of Public Economics 93: 1058–1068. Dyck, A., and L. Zingales. 2004. Private benefits of control: An international comparison. The Journal of Finance LIX (2): 537–600. Dyreng, S., M. Hanlon, and E. L. Maydew. 2008. Long-run corporate tax avoidance. The Accounting Review 83: 61–82. El Ghoul, S., O. Guedhami, and J. Pittman. 2011. The role of IRS monitoring in equity pricing in public firms. Contemporary Accounting Research 28: 643–674. Erickson, M., M. Hanlon and E. L. Maydew. 2004. How much will firms pay for earnings that do not exist? Evidence of taxes paid on allegedly fraudulent earnings. The Accounting Review 79 (2): 387-408. Estrin, S., and M. Prevezer. 2011. The role of informal institutions in corporate governance: Brazil, Russia, India, and China compared. Asia-Pacific Journal of Management 28: 41–67. Faccio, M. 2006. Politically connected firms. American Economic Review 96 (1): 369–386. Fan, G., X. L. Wang, and H. Zhou. 2011. NERI Index of Marketization of China’s provinces. Beijing, PRC: The Economic Science Press (in Chinese). Fan, J. P. H., T. J. Wong, and T. Zhang. 2007. Politically-connected CEOs, corporate governance and post-IPO performance of China’s newly partially privatized firms. Journal of Financial Economics 84: 330–357. Fisman, R. 2001. Estimating the value of political connections. American Economic Review 91 (4): 1095–1102. Francis, B. B., I. Hasan, and X. Sun. 2012. CEO political affiliation and firms’ tax avoidance. Available at http://ssrn.com/abstract=2013248. Frank, M. M., L. J. Lynch, and S. O. Rego. 2009. Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review 84 (2): 467–496. Greene, W. 2008. Econometric Analysis. Upper Saddle River, NJ: Pearson/Prentice Hall. Guedhami, O., and J. Pittman. 2008. The importance of IRS monitoring to debt pricing in private firms. Journal of Financial Economics 90: 38–58. Guedhami, O., J. A. Pittman, and W. Saffar. 2014. Auditor choice in politically connected firms. Journal of Accounting Research 52 (1): 107–162. Guenther, D., and D. Young. 2000. The association between financial accounting measures and real economic activity: A multinational study. Journal of Accounting and Economics 29 (1): 53–72. Gupta, S., and L. F. Mills. 2002. Corporate multistate tax planning: Benefits of multiple jurisdictions. Journal of Accounting and Economics 33: 117–139. This article is protected by copyright. All rights reserved. Hanlon, M., and S. Heitzman. 2010. A review of tax research. Journal of Accounting and Economics 50 (2–3): 127–178. Hanlon, M., J. L. Hoopes, and N. Shroff. 2014. The effect of tax authority monitoring and enforcement on financial reporting quality. Journal of the American Taxation Association 36 (2): 137–170. Haveman, H. A., N. Jia, J. Shi, and Y. Wang. 2017. The dynamics of political embeddedness in China. Administrative Science Quarterly, 62 (1): 67-104. Haw, I. M., B. Hu, L. S. Hwang, and W. Wu. 2004. Ultimate ownership, income management, and legal and extra-legal institutions. Journal of Accounting Research 42 (2): 423–462. He, X., T. J. Wong, and D. Young. 2012. Challenges for implementation of fair value accounting in emerging markets: Evidence from China. Contemporary Accounting Research 29 (2), 538–562. Heckman, J., H. Ichimura, and P. Todd. 1998. Matching as an econometric evaluation estimator. Review of Economic Studies 65: 261–94. Hoopes, J. L., D. Mescall, and J. A. Pittman. 2012. Do IRS audits deter corporate tax avoidance? The Accounting Review 87 (5): 1603–1639. Houston, J., L. Jiang, C. Lin, and Y. Ma. 2014. Political connections and the cost of bank loans. Journal of Accounting Research 52 (1): 193–243. Huang, J., and C. Zhao. 2015. Domestic transfer pricing: Two firms paid an extra RMB5 million in taxes due to failed tax planning. China Tax News. 25 March 2015. Jacob, J. 1996. Taxes and transfer pricing: Income shifting and the volume of intrafirm transfers. Journal of Accounting Research 34 (2): 301–312. Johnson, S., and T. Mitton. 2003. Cronyism and capital controls: Evidence from Malaysia. Journal of Financial Economics 66, 351–382. Judge, G. G., R. C. Hill, W. E. Griffiths, W. E. Lutkepohl, and T. C. Lee. 1988. Introduction to the Theory and Practice of Econometrics, 2nd Edition. New York: John Wiley and Sons, Inc. Khwaja, A. I., and A. Mian. 2005. Do lenders favor politically connected firms? Rent provision in an emerging financial market. The Quarterly Journal of Economics 120 (4): 1371-1411. Kim, C. F., and L. Zhang. 2016. Corporate political connections and tax aggressiveness. Contemporary Accounting Research 33 (1): 78-114. Larcker, D. F., and T. O. Rusticus. 2010. On the use of instrumental variables in accounting research. Journal of Accounting and Economics 49: 186–205. Lennox, C, J. R. Francis, and Z. Wang. 2012. Selection models in accounting research. The Accounting Review 87: 589–616. Lennox, C., W. Li, J. Pittman, and Z. Wang. 2015. The determinants and consequences of tax audits: Some evidence from China. Working paper at SSRN: http://ssrn.com/abstract=2585466. Levine, R. 2005. Law, Endowments and Property Rights. Journal of Economic Perspectives 19 (3): 61-88. Li, H., L. Meng, and J. Zhang. 2006. Why do entrepreneurs enter politics? Economic Inquiry 44 (3): 559–578. Li, H., and L. Zhou. 2005. Political turnover and economic performance: The disciplinary role of personnel control in China. Journal of Public Economics 89: 1743–1762. Lin, B., R. Lu, and T. Zhang. 2012. Tax-induced earnings management in emerging markets: Evidence from China. Journal of the American Taxation Association (Fall): 19–44. Lin, K. Z., L. F. Mills, and F. Zhang. 2014. Public versus private firm responses to the tax rate reduction in China. Journal of the American Taxation Association 36 (1). Liu, Q., W. Luo, and P. Rao. 2015. The political economy of corporate tax avoidance. This article is protected by copyright. All rights reserved. Working paper at SSRN: http://ssrn.com/abstract=2709608. Lo, A. W., R. M. Wong, and M. Firth. 2010. Tax, financial reporting, and tunneling incentives for income shifting: An empirical analysis of the transfer pricing behavior of Chinese-listed companies. Journal of the American Taxation Association (Fall): 1–26. Loughran, T. 2008. The impact of firm location on equity issuance. Financial Management 37: 1–21. Markle, K. 2016. A comparison of the tax-motivated income shifting of multinationals in territorial and worldwide countries. Contemporary Accounting Research 33 (1): 7-43. Mills, L. F. 1998. Book-tax differences and internal revenue service adjustments. Journal of Accounting Research 36: 343–356. Mills, L. F., S. E. Nutter, and C. M. Schwab. 2013. The effect of political sensitivity and bargaining power on taxes: Evidence from federal contractors. The Accounting Review 88 (3): 977–1005. Moses, R. L. 2015. The Chinese government has a resignation problem. The Wall Street Journal, 24 April 2015. Petersen, M. 2009. Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies 22 (1): 435–480. Piotroski, J. D., and T. J. Wong. 2012. Institutions and information environment of Chinese-listed firms. In Capitalizing China, edited by J. Fan and R. Morck. Chicago: National Bureau of Economic Research, University of Chicago Press. PricewaterhouseCoopers. 2012. More lenient treatments for certain corporate income tax deduction issues. China Tax and Business News Flash 13, May 2012. PricewaterhouseCoopers. 2015. Localization of BEPS actions is moving ahead in China. China Tax and Business News Flash 31, July 2015. Shevlin, T., T. Tang, and R. Wilson. 2012. Domestic income shifting by Chinese listed firms. Journal of the American Taxation Association 34 (1): 1–29. State Administration of Taxation (SAT) Inspection Bureau. 2004–2008. China Tax Audits Yearbook. Beijing: China Tax Press Pub. State Administration of Taxation (SAT). 2009. Implementation measures of special tax adjustments--Trial version, Guo Shui Fa  Circular No. 2. Towery, E. 2017. Unintended consequences of linking tax return disclosures to financial reporting for income taxes: Evidence from Schedule UTP. The Accounting Review In-Press. Vuong, Q. H. 1989. Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 57: 307-333. World Bank. 2006. Governance, investment climate, and harmonious society: Competitiveness enhancements for 120 cities in China. Report No. 37759-CN, October 8, 2006. Wu, W., S. A. Johan, and O. M. Rui. 2016. Institutional investors, political connections, and the incidence of regulatory enforcement against corporate fraud. Journal of Business Ethics 134: 709–726. Wu, L., Y. Wang, B. X. Lin, C. Li, and S. Chen. 2007. Local tax rebates, corporate tax burdens, and firm migration: Evidence from China. Journal of Accounting and Public Policy 26(5): 555–583. Xin, K. R., and J. L. Pearce. 1996. Guanxi: Connections as substitutes for formal institutional support. Academy of Management Journal 39 (6): 1641–1658. Young, M., M. Reksulak, and W.F. Shughart. 2001. The political economy of the IRS. Economics & Politics 13 (2): 201–220. This article is protected by copyright. All rights reserved. Figure 1 Mean tax enforcement for each province (2003–2013) Notes: To help readers understand the enforcement intensity across geographic locations, we use four shades to show how many provinces fall into each quartile of Enforcement for the period 2003–2013. Provinces in the darkest shade are those with the highest enforcement level. Firms in these regions are most likely to be audited, confront a well-trained and skilled audit workforce, and face the toughest fines for underreporting tax income. This article is protected by copyright. All rights reserved. Appendix 1 Principal component analysis proxies for tax enforcement We use aggregate data from the China Tax Audits Yearbook (2003–2013) to construct our tax enforcement measures. The yearbook is published annually by the SAT and contains detailed tax enforcement information for each provincial (SAT and local) tax bureau in China. Specifically, the disclosed information includes the quantity of (i) permanent employees, (ii) tax inspectors, and (iii) employees with a bachelor’s degree and/or a professional qualification (certified accountants, certified tax agents, and lawyers) and their Communist Party membership and age range; (iv) corporate taxpayers; (v) corporate tax returns audited; (vi) audit departments; (vii) suspicious cases; (viii) cases prosecuted; (ix) cases closed; (x) regional tax revenue; (xi) tax deficiencies settled; (xii) overdue tax surcharges; and (xiii) tax penalties, interest, and fines. If information about a specific item is missing for a certain year, we replace the missing information with data from earlier or later years if it is not available from other sources. To extract common factors from a pooled data set, we use principal component analysis, a tool used to analyze exploratory (rather than confirmatory) data and adopted in many prior studies (Dechow et al. 1996; Bushee 1998; Kim and Zhang 2016). We construct nine ratios using the data described above. Specifically, Ratio 1 and Ratio 2 are the audit coverage and the prosecution rate, respectively, while Ratio 3 captures the manpower sufficiency to monitor listed firms. Ratio 4 to Ratio 7 reflect the educational and professional level of the tax officers in a region, and Ratio 8 and Ratio 9 measure the economic consequences of the tax audits in terms of additional tax payments and monetary punishment for tax underreporting. To mitigate the difficulty of drawing conclusions from one ratio at a time and to alleviate the multicollinearity concern of using multiple ratios in the same regression, we perform principal component analysis with an oblique (rather than an orthogonal) rotation to identify and interpret common factors. This procedure reduces the dimensionality of the data by combining the nine individual aspects into a parsimonious set of linear combinations that explain the shared variance among the characteristics. Our analysis produces three common factors with an eigenvalue greater than one. These three factors retain 69 percent of the total variance of the original data. For ease of interpretation, we associate each factor with those variables that have post-rotation loadings larger than 0.60 in absolute value and are statistically different from zero at conventional levels. Factor 1 is associated with the number of tax audits divided by the number of tax returns filed in the same region (factor loading = 0.917) and the number of cases prosecuted divided by the number of tax returns filed (factor loading = 0.933). Clearly, Factor 1 represents the probability of facing a tax audit and the likelihood of being prosecuted in a region-year. Labeled Probability, this factor explains the largest proportion (33 percent or three units) of the total variance. Factor 2 is associated with the percentage of tax officers holding the qualification of certified public accountant (factor loading = 0.877), certified tax agent (factor loading = 0.848), or lawyer (factor loading 0.716). This factor, which we label Expertise, measures the proficiency of tax officers and explains 22 percent of the total variance. Higher percentages represent higher-quality audit personnel in the region, and, plausibly, a greater likelihood that tax misreporting will be detected. Factor 3 is associated with the dollar value of tax deficiencies settled (factor loading = 0.906) and the dollar value of surcharges, penalties, and fines (factor loading = 0.876), respectively, over the amount of regional tax revenue. This factor, labeled Outcome, measures the outcomes or consequences of tax audits and explains 14 percent of the total variance. As the remaining factors account for only trivial amounts of variance, they are not retained. To form an aggregate enforcement measure, which we label Enforcement, we first rank each factor into percentile groups and then calculate the mean ranking of the three factors for each region and year (2003–2013). A higher aggregate value indicates higher enforcement intensity. This article is protected by copyright. All rights reserved. Appendix 2 Variable definitions and construction Dependent variable ETR GAAP effective tax rate, annual income tax expense scaled by annual pre-tax income before special items Explanatory variables Tax enforcement measures at the regional-year level: Input variables Ratio 1 Number of corporate income tax audits conducted over the number of corporate income tax returns filed Ratio 2 Number of income tax cases prosecuted over the number of income tax returns filed Ratio 3 Number of tax officers over the number of listed firms Ratio 4 Number of tax officers with a bachelor’s degree or higher over the number of tax officers Ratio 5 Number of tax officers with a CPA qualification over the number of tax officers Ratio 6 Number of tax officers with a certified tax agent qualification over the number of tax officers Ratio 7 Number of tax officers with a lawyer qualification over the number of tax officers Ratio 8 Amount of tax deficiencies settled over the amount of regional tax revenues Ratio 9 Amount of penalties, interest, and fines over the amount of regional tax revenues Factor extraction and variables used in regression analysis Factor 1 Probability of tax audits, measured by ratios 1, 2, and 3 Factor 2 Expertise of tax officers, measured by ratios 4 to 7 Factor 3 Outcome or consequence of tax audits, measured by ratios 8 and 9 Enforcement Average percentile ranking of factors 1 to 3 Political connection measures at the firm-year level: ConnectedChair One if the chairman of the board of directors is politically connected and zero otherwise ConnectedBoard One if the number of politically connected board members is above the industry-year median and zero otherwise Connected# Natural logarithm of one plus the number of connected members on board Connected% Percentage of connected board members on board Control variables ROA Ratio of consolidated net income to consolidated total assets Std. dev. of ROA Standard deviation of ROA Size Natural logarithm of year-end total assets Liquidity Ratio of current assets to current liabilities Leverage Ratio of total debts to total assets PPE Ratio of property, plant, and equipment to total assets Growth Ratio of market value to book value of net assets Intangible Ratio of intangible assets to total assets Inventory Ratio of inventory to total assets Cash Ratio of year-end cash holdings to lagged assets Accruals Discretionary current accruals estimated from the modified cross-sectional Jones model Ownership Percentage of shares owned by the government GDP Natural logarithm of provincial annual GDP Institution Index that reflects institutional characteristics of a province, based on the National Economic Research Institute (NERI) Index of Marketization of China’s provinces LeaderTurnover One if the province where the firm is headquartered experiences a governor or party secretary turnover and zero otherwise Industry Set of dummy variables indicating industry sector membership based on the CSRC classification Year Set of dummy variables indicating years This article is protected by copyright. All rights reserved. TABLE 1 Sample distribution by year and industry 2003 2004 A. Agriculture 13 13 B. Natural resources 2005 16 2006 17 2007 16 2008 17 2009 15 2010 21 2011 16 2012 21 2013 17 All 182 % 1.637 23 27 24 26 34 34 40 42 44 45 46 385 3.462 482 504 498 520 538 454 488 543 538 535 558 5,658 D. Utilities 52 52 56 55 60 43 60 60 57 61 67 623 50.87 7 5.602 E. Construction 12 15 18 20 20 19 21 23 24 26 29 227 2.041 F. Transportation 42 45 47 46 55 50 53 58 61 61 60 578 5.197 G. Information technology 43 47 38 49 49 39 47 52 56 55 55 530 4.766 H. Wholesale and retail 82 86 83 85 91 82 85 89 90 92 95 960 8.632 J. Real estate 84 78 74 83 91 88 103 104 108 106 110 1,029 9.253 K. Services 31 34 28 31 35 32 29 36 35 38 42 371 3.336 L. Communication 14 14 10 9 14 11 10 16 17 20 21 156 1.403 M. Conglomerates 42 37 33 36 38 34 38 37 44 41 42 422 3.795 920 952 925 977 1,041 903 989 1,08 1 1,090 1,101 1,142 11,121 100 C. Manufacturing Total Notes: This table presents the sample distribution by year and industry following the 13 CSRC industry classifications. This article is protected by copyright. All rights reserved. TABLE 2 Descriptive statistics of main variables Mean P25 Median P75 Std. dev. ETR 0.233 0.135 0.205 0.293 0.158 Tax enforcement measures Input variables Ratio 1 0.016 0.003 0.009 0.021 0.020 Ratio 2 0.012 0.003 0.005 0.013 0.016 Ratio 3 16.863 6.043 12.500 22.917 13.682 Ratio 4 0.924 0.903 0.935 0.958 0.049 Ratio 5 0.015 0.009 0.014 0.020 0.009 Ratio 6 0.059 0.038 0.046 0.076 0.038 Ratio 7 0.006 0.002 0.004 0.008 0.007 Ratio 8 0.013 0.006 0.010 0.017 0.011 Ratio 9 0.002 0.001 0.001 0.002 0.001 Factor extraction and variables used in regression analysis Audit Probability (Factor 1) -0.168 -0.785 -0.486 0.137 0.925 Auditor Expertise (Factor 2) 0.547 -0.159 0.403 1.087 1.107 Audit Outcome (Factor 3) -0.023 -0.623 -0.304 0.308 0.940 Enforcement (All factors) 0.511 0.372 0.517 0.606 0.152 Political connection measures ConnectedChair 0.231 0.000 0.000 0.000 0.421 ConnectedBoard 0.343 0.000 0.000 1.000 0.475 Connected# 0.064 0.000 0.693 1.099 0.559 Connected% 8.749 0.000 7.143 13.333 9.244 Notes: The sample statistics are based on 11,121 firm-year observations from 2003 to 2013. See Appendix 2 for variable definitions and construction. This article is protected by copyright. All rights reserved. TABLE 3 Univariate tests All firms Connected Nonconnected Panel A: Comparison of the mean ETR between connected and nonconnected firms 1st quartile Enforcement (weakest) 0.225 0.224 0.225 2nd quartile Enforcement 0.232 0.228 0.231 3rd quartile Enforcement 0.238 0.231 0.243 4th quartile Enforcement (strongest) 0.238 0.230 0.245 Difference Panel B: Comparison of the mean tax enforcement between connected and nonconnected firms Audit Probability -0.168 -0.183 -0.160 Auditor Expertise 0.547 0.550 0.546 Audit Outcome -0.023 -0.043 -0.013 Enforcement (overall) 0.511 0.512 0.510 -0.001 -0.003 -0.012* -0.015** -0.023 0.004 -0.031 0.002 Panel C: Comparison of the mean firm characteristics between connected and nonconnected firms ROA 0.045 0.049 0.042 0.007*** Std. dev. of ROA 0.030 0.028 0.031 -0.003** Size 21.890 22.258 21.699 0.558*** Liquidity 1.568 1.572 1.566 0.006 Leverage 0.504 0.508 0.502 0.006 PPE 0.275 0.269 0.279 -0.009** Intangible 0.043 0.043 0.044 -0.001 Growth 3.171 3.031 3.243 -0.213*** Inventory 0.182 0.187 0.180 0.007** Cash 0.160 0.163 0.159 0.005** Accruals 0.012 0.016 0.010 0.006*** Ownership 0.206 0.198 0.211 -0.012** GDP 9.454 9.444 9.460 -0.016 Institution 7.547 7.588 7.526 0.061 LeaderTurnover 0.358 0.361 0.354 0.007 N 11,121 3,810 7,311 Notes: This table shows the differences between the two sample means for ETR, tax enforcement, and firm characteristics and the two-sample t-test results. Politically connected firm-years are defined as those firm-years that the number of politically connected board members is above the industry-year median. See Appendix 2 for variable definitions and construction. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels, respectively. This article is protected by copyright. All rights reserved. TABLE 4 The effect of tax enforcement on firms’ ETR (1) Audit Probability 0.006** (2.16) Auditor Expertise Audit Outcome (2) (3) 0.004* (1.94) 0.009*** (3.22) (4) 0.006** (2.06) 0.004* (1.87) 0.009*** (3.21) Enforcement ROA St. dev. of ROA Size Liquidity Leverage PPE Intangibles Growth Inventory Cash Accruals Ownership GDP Institution LeaderTurnover Intercept Year fixed effects Industry fixed effects N Adj. R2 F-statistic -1.247*** (-18.03) -0.051 (-1.14) -0.003 (-1.18) 0.001 (0.46) 0.011 (0.57) 0.060*** (3.74) 0.099*** (2.64) 0.003*** (2.85) 0.127*** (5.16) 0.106*** (4.50) -0.057*** (-3.55) -0.006 (-0.68) 0.023*** (5.24) -0.005*** (-2.63) -0.003 (-0.85) -0.015 (-0.25) Yes Yes 11,121 0.158 22.62 -1.254*** (-18.13) -0.048 (-1.07) -0.003 (-1.07) 0.001 (0.40) 0.009 (0.47) 0.058*** (3.66) 0.096** (2.56) 0.003*** (2.95) 0.127*** (5.14) 0.103*** (4.38) -0.056*** (-3.53) -0.005 (-0.56) 0.023*** (5.27) -0.007*** (-3.18) -0.002 (-0.76) -0.008 (-0.14) Yes Yes 11,121 0.158 22.98 This article is protected by copyright. All rights reserved. -1.251*** (-18.04) -0.050 (-1.11) -0.002 (-0.87) 0.001 (0.35) 0.009 (0.43) 0.058*** (3.62) 0.090** (2.40) 0.003*** (2.97) 0.126*** (5.14) 0.105*** (4.47) -0.057*** (-3.58) -0.002 (-0.23) 0.021*** (4.85) -0.004* (-1.76) -0.002 (-0.58) -0.030 (-0.51) Yes Yes 11,121 0.159 22.89 -1.250*** (-18.07) -0.049 (-1.08) -0.002 (-0.84) 0.001 (0.48) 0.010 (0.50) 0.058*** (3.68) 0.092** (2.47) 0.003*** (3.06) 0.128*** (5.24) 0.105*** (4.49) -0.056*** (-3.52) -0.003 (-0.34) 0.022*** (5.10) -0.004** (-2.15) -0.002 (-0.76) -0.040 (-0.68) Yes Yes 11,121 0.160 22.21 (5) 0.047*** (3.43) -1.249*** (-18.05) -0.049 (-1.10) -0.002 (-0.96) 0.001 (0.45) 0.010 (0.48) 0.058*** (3.64) 0.096** (2.56) 0.003*** (2.96) 0.128*** (5.22) 0.105*** (4.48) -0.056*** (-3.51) -0.005 (-0.58) 0.022*** (5.05) -0.005*** (-2.64) -0.002 (-0.75) -0.014 (-0.24) Yes Yes 11,121 0.159 23.01 Notes: This table reports the regression results for the effect of aggregate and dimensional tax enforcement on firms’ ETR. t-statistics are reported in parentheses below the coefficient. All variables are defined in Appendix 2. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels (two-tailed), respectively. This article is protected by copyright. All rights reserved. TABLE 5 The effect of tax enforcement on firms’ ETR conditional on political connectedness ConnectedChair ConnectedBoard Connected# (1) (2) (3) *** *** Enforcement (β1) 0.059 0.066 0.076*** (3.71) (3.88) (3.46) Connected (β2) 0.002 0.004 0.005 (0.33) (0.83) (1.22) -0.051* -0.058** -0.044* Enforcement×Connected (β3) (-1.69) (-2.17) (-1.85) ROA -1.251*** -1.250*** -1.251*** (-18.07) (-18.01) (-18.01) -0.051 -0.051 -0.050 St. dev. of ROA (-1.12) (-1.12) (-1.11) Size -0.002 -0.003 -0.003 (-1.05) (-1.16) (-1.33) Liquidity 0.001 0.001 0.001 (0.49) (0.48) (0.51) Leverage 0.010 0.011 0.012 (0.51) (0.55) (0.61) PPE 0.059*** 0.060*** 0.060*** (3.70) (3.76) (3.75) Intangibles 0.096** 0.096** 0.095** (2.55) (2.56) (2.55) Growth 0.003*** 0.003*** 0.003*** (2.97) (2.93) (2.89) *** *** Inventory 0.129 0.130 0.129*** (5.25) (5.30) (5.27) Cash 0.105*** 0.105*** 0.105*** (4.50) (4.50) (4.50) Accruals -0.056*** -0.057*** -0.057*** (-3.54) (-3.57) (-3.60) Ownership -0.005 -0.005 -0.005 (-0.49) (-0.52) (-0.49) GDP 0.022*** 0.023*** 0.023*** (5.11) (5.16) (5.15) Institution -0.005*** -0.005*** -0.005*** (-2.70) (-2.73) (-2.75) LeaderTurnover -0.002 -0.002 -0.002 (-0.77) (-0.78) (-0.78) Intercept -0.012 -0.009 -0.002 (-0.20) (-0.15) (-0.03) Year and industry fixed effects Yes Yes Yes N 11,121 11,121 11,121 Adj. R2 0.159 0.160 0.160 F-statistic 22.09 22.11 22.02 Test of joint significance (columns 1 and 2) Coeff. F-stat Coeff. F-stat β 1 + β3 = 0 0.008 0.10 0.008 0.14 This article is protected by copyright. All rights reserved. Connected% (4) 0.068*** (3.53) 0.000 (1.61) -0.002* (-1.69) -1.252*** (-18.04) -0.050 (-1.11) -0.003 (-1.35) 0.001 (0.52) 0.012 (0.62) 0.060*** (3.77) 0.095** (2.55) 0.003*** (2.87) 0.129*** (5.27) 0.105*** (4.50) -0.058*** (-3.62) -0.004 (-0.41) 0.023*** (5.18) -0.005*** (-2.77) -0.002 (-0.79) -0.002 (-0.04) Yes 11,121 0.160 22.05 Notes: This table reports the regression results for the (aggregate) tax enforcement effect on firms’ ETR conditional on different measures of political connections. t-statistics are reported in parentheses. All variables are defined in Appendix 2. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels (two-tailed), respectively. This article is protected by copyright. All rights reserved. TABLE 6 The effect of changes in political connections Panel A1: The effect of gaining a new connection Enforcement (β1) 0.076*** (2.76) Treatment (β2) 0.001 (0.11) Post (β3) 0.003 (0.39) -0.019 Enforcement×Treatment (β4) (-0.49) 0.002 Enforcement×Post (β5) (0.05) -0.016* Treatment×Post (β6) (-1.83) -0.111 Enforcement×Treatment×Post (β7) (-1.85)* Control variables Yes Year fixed effects Yes Industry fixed effects Yes N 3,656 Adj. R2 0.128 F-statistic 15.96 Panel A2: Tests of coefficient differences Pre-period Post-period Diff. β1 β1+β5 β5 Control firms Enforcement×ConnectedBoard×Post2011 (β7) Control variables Year fixed effects Industry fixed effects N Adj. R2 F-statistic Panel B2: Tests of coefficient differences Pre-2011 Post-2011 β1 β1+β5 Unconnected firms Treatment firms Connected firms Diff. = 0.076*** β1+β4 = 0.057** β4 = -0.019 = 0.078** β1+β4+β5+β7 = -0.052 β4+β7 = -0.130*** = 0.002 β5+β7 = -0.109*** β7 = -0.111* Panel B1: The effect of a regime change Enforcement (β1) ConnectedBoard (β2) Post2011 (β3) Enforcement×ConnectedBoard (β4) Enforcement×Post2011 (β5) ConnectedBoard×Post2011 (β6) Diff. 0.053* (1.76) -0.003 (-0.35) 0.017** * (3.10) -0.120* ** (-2.61) -0.061 (-1.53) 0.006 (0.70) 0.166** (2.41) Yes No Yes 4,414 0.186 14.916 Diff. β5 = 0.053* = -0.008 = -0.061 β1+β4 β1+β4+β5+β7 β5+β7 = -0.067* = 0.038 = 0.105* β4 β4+β7 β7 = -0.120*** = 0.046 = 0.166** Notes: This table reports the regression results for the effect of changes in political connections. The same set of control variables as in Table 5 is included in both regressions, but the results are not reported here for brevity. Panels A1 and B1, respectively, show the results on the effect of obtaining a new connection and changing political landscape. Panels A2 and B2, respectively, show the coefficient differences between the two groups of firms across time. t-statistics are reported in parentheses. Treatment is an indicator variable that equals one for firms that increase their political connections, and zero otherwise. Post equals one for firm-years in the post-change period, and zero otherwise. Post2011 equals one for the 2012–2013 observations and zero for the 2010–2011 observations. Other variables are defined in Appendix 2. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels (two-tailed), respectively. This article is protected by copyright. All rights reserved. TABLE 7 Heckman two-step treatment effects procedure to correct for endogeneity First Stage DV = ConnectedBoard Enforcement 0.218 (1.32) ConnectedBoard Enforcement×ConnectedBoard Inverse Mills Ratio Industry % of Connected Firms Intercept 3.043*** (7.82) -6.002*** (-7.87) Yes Yes Yes 11,121 0.057 217.7 Second Stage DV = ETR 0.069*** (4.01) 0.004 (0.87) -0.058** (-2.18) 0.015 (0.79) -0.074 (-0.76) Yes Yes Yes 11,121 0.160 21.70 Control variables Year fixed effects Industry fixed effects N Pseudo/Adjusted R2 Wald χ2 Notes: This table presents the regression results of the Heckman two-step procedure. The same set of control variables as in Table 5 is included in both regressions, but the results are not reported here for brevity. t-statistics are reported in parentheses. Industry % of Connected Firms is defined as the percentage of firms with a politically connected board in a firm’s industry group. Other variables are defined in Appendix 2. *** and ** indicate significance at the 0.01 and 0.05 levels (two-tailed), respectively. This article is protected by copyright. All rights reserved. TABLE 8 Panel A: Joint effect of political connections and tax enforcement on income shifting (1) (2) Shifting (β1) -0.063*** -0.081*** (-3.13) (-4.07) Enforcement (β2) 0.007* (2.08) ConnectedBoard (β3) Shifting×Enforcement (β4) Shifting×ConnectedBoard (β5) Enforcement×ConnectedBoard (β6) Shifting×Enforcement×ConnectedBoard (β7) Intercept 0.072*** (3.50) (3) -0.068** (-2.69) 0.012* (1.89) 0.007* (1.90) 0.143*** (3.21) -0.035 (-0.92) -0.013 (-1.23) -0.125* (-1.86) 0.086 (1.68) Yes 7,869 0.155 16.514 0.088 0.083 (1.73) (1.61) Control variables Yes Yes N 7,869 7,869 Adj. R2 0.154 0.154 F-statistic 18.172 17.690 Panel B: Tests of coefficient differences between groups Weak enforcement Strong enforcement Diff. Unconnected firms β1 β1 + β4 β4 = -0.068** = 0.075 = 0.143*** Connected firms β1 + β5 β1 + β4 + β5 + β7 β4 + β7 *** *** = -0.103 = -0.085 = 0.018 Diff. β5 β5 + β7 β7 = -0.035 = -0.160*** = -0.125* Notes: This table reports the regression results of the joint effect of political connections and tax enforcement on income shifting (panel A) and the test for the coefficient differences between connected and unconnected firms (panel B). The same set of control variables as in Table 5 is included in all the regressions but not reported here for brevity. t-statistics are reported in parentheses. Enforcement is an indicator variable that equals one if the province’s estimated Enforcement value is in the top quartile in that year, and zero otherwise. Other variables are defined in Appendix 2. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels (two-tailed), respectively. This article is protected by copyright. All rights reserved.