Many show that left-leaning governments tend to maintain a high statutory tax rate (STR), the corporate tax rate codified in law, in keeping with their political views (Inclan, Quinn, and Shapiro 2001; Osterloh and Debus 2012; Quinn and Shapiro 1991a). Despite the strong finding on the positive relationship between left-leaning governments and high corporate tax rates in political science scholarship, business scholars have demonstrated that firms face widely differing effective tax rates (ETRs), actual tax payments firms pay. For example, the United States had a 35 percent corporate STR in 2012¹, but large corporations in the United States paid an average ETR of 12.6 percent, as differing firm characteristics and lobbying activities can lower their ETR (Gupta and Newberry 1997; Richter, Samphantharak, and Timmons 2009; Vandenbussche, Crabb, and Janssen 2005; Zimmerman 1983). Tax incentives, “provisions of tax law, regulation, or practice that reduce or postpone revenue for a comparatively narrow population of taxpayers relative to a benchmark tax,” (OECD 2010), play a primary role in determining ETR (Martin 1991; Steinmo 1993).² Thus the relatively high corporate STRs imposed by left-leaning governments may mask much lower ETRs.

In this article, I explain these puzzling patterns by focusing on the policy instruments that politicians utilize to maximize their electoral success. In particular, I shed light on tax incentives as a main cause for varying ETRs at the firm level. I argue that left-oriented politicians tend to keep all corporate tax rates high but provide generous tax incentives to compensate for the high corporate tax rates. Constituency costs influence political parties' incentives to employ opaque policy instruments. This constituency cost varies by the level of policy complexity. Although voters can evaluate the effects of a transparent policy instrument, such as the corporate tax rate, on their personal welfare, assessing the impact of more opaque policy instruments, such as tax incentives, is more difficult. Given this, transparent policy instruments will have higher constituency costs and benefits than non-transparent policy instruments. In turn, political parties have incentives to choose more opaque policy instruments to minimize constituency costs and maximize their electoral benefits. This effect is much stronger for left-wing governments because they must balance the need to attract campaign donations from firms and promote economic growth with the need to garner voter support through possibly increasing corporate taxes.

To test my argument, I employ a Bayesian Item Response Theory (IRT) model to measure ideal points of Senators’ voting and logistic regression analyses using tax legislation on tax rates and tax incentives in the United States between 1986 and 2012. The findings support my hypothesis: Democrats tend to oppose tax rate reductions but favor generous tax incentives intended to develop the manufacturing industry. This finding also shows the importance of the Democratic party's alignment with labor. Consistent with the ideal points measured, Democrats and legislators representing the interests of labor are more likely to oppose reductions in tax rates but favor generous tax incentives for the manufacturing industry.

In what follows, I describe relevant literature and present theoretical arguments. The research design section introduces data, variables, and methods. The results section discusses empirical results. The final section concludes and discusses the implications of my findings.

While political scientists generally agree that domestic political institutions play a pivotal role in corporate tax policy, they have reached little agreement about how. Some find a correlation between left-wing power and higher corporate tax rates because of redistributive mechanisms and growth strategies (Inclan et al. 2001; Osterloh and Debus 2012). For instance, Basinger and Hallerberg (2004) argue that left-wing governments are reluctant to implement corporate tax reforms to respond to tax competition because of constituency costs. These findings accord with the broader understanding of left-wing governments as redistributing wealth and increasing public spending to benefit low-skilled workers, their core constituents (Osterloh and Debus 2012; Quinn and Shapiro 1991a). Leftist governments are also more likely to pursue consumption-driven economic growth strategies associated with a high capital tax rate and lower rates of interest while right-wing governments tend to undertake investment-driven economic growth strategies accompanied by high rates of saving and lower rates of capital taxation (Boix 1998; Inclan et al. 2001; Quinn and Shapiro 1991a). Recent scholarship demonstrates that government partisanship continues to influence fiscal policy. A study found that individuals who identify with the same political party as the incumbent president express fewer negative views on government tax and spending policies (Cullen et al. 2021). Additionally, counties with Democratic legislators tend to spend more than those with Republican legislators (de Benedictis-Kessner and Warshaw 2020). While this body of literature provides important insights, it does not address variations in firms’ ETR, which, can be quite wide. The domestic institutional approach in political science has overlooked the role of tax incentives in determining corporate ETR.

Unlike political science scholarship, economics and business scholars have explained differences in ETRs among firms by focusing on firm characteristics –size, type of assets, capital intensity, and foreign ownership levels.³ Others have pointed to the role of firms' lobbying activities for policy measures such as tax credits for investment, preferential tax rates, and favorable depreciation rules as determinants of ETR (Richter et al. 2009). Yet, these firm-level explanations treat the government as a passive actor, which does not fully explain the wide variation in firms' ETRs. One notable exception is Pinto (2013), who explores government partisanship and foreign direct investment. Pinto (2013) suggests that left-wing governments provide lower tax rates to multinational corporations (MNCs) primarily to attract foreign direct investment (FDI) inflows and that right-wing governments provide higher tax rates to internationally mobile firms to protect domestic firms. Thus, he concludes that left-wing governments lower tax rates for internationally mobile firms while imposing high tax rates on immobile firms (domestic firms).

While I concur with Pinto (2013) that the political leanings of the governing party may play a role in corporate ETRs, I do not attribute this difference solely to the approach to attracting FDI. In arguing that constituency costs compel left-wing governments to employ more opaque policies than right-wing governments, I focus primarily on influence of partisan governments on policy transparency, placing such effects within corporate tax policy. Pintós main concern is about the impact of partisan government on FDI flows; preferential tax rates for multinational corporations are one of the various measures that benefit MNCs. I also use more direct data – corporate tax legislation. This allows me to examine the primary causal mechanism by which the left provides tax benefits or other favorable conditions to firms. Aggregate-level data, such as the amount of FDI flows at both national and sector levels, provides a less direct analysis of partisanship's effects.

In this study, I focus on the role of tax incentives, which have a large impact on ETR and ultimately determine the wide variation in ETRs across firms within a country (Martin 1991; Steinmo 1993). By focusing on a wide range of policy instruments that legislators can use to set corporate tax policy, I can offer a more complete explanation for why left-wing governments have higher STRs as found in the existing corporate tax policy literature and simultaneously have lower ETRs and a wider variation in ETRs across firms. If we consider the obscure policy instruments that determine the differences in firms' ETRs, the net impact of partisan governments on corporate tax policy is likely to be ambiguous. In the next section, I propose a more complex relationship between government partisanship and corporate tax policy outcomes.

Elsewhere, I have argued that left-leaning governments adopt more opaque policy tools than their right-leaning counterparts (Shin 2017), drawing upon the literature regarding constituency costs in corporate tax policy (Basinger and Hallerberg 2004) and on the literature related to transparency in trade policy (Kono 2006). Here, I extend this theory to explain how different political parties design corporate tax policy in a distinctive way by focusing on a policy tool, tax incentives.

First, the corporate tax policy literature shows that constituency costs hinder corporate tax reforms. (Basinger and Hallerberg 2004). Swank (2002) defines constituency costs as “governing parties' potential opposition to tax reform or its policy repercussions.” Basinger and Hallerberg (2004) separate domestic political costs into transaction costs – the costs of moving any piece of legislation through the legislative process – and constituency costs – the costs of potential opposition to tax cuts or particular tax reforms by pivotal actors within the legislative process. They define constituency costs, my focus here, as “the costs of ideological opposition to tax cuts or particular tax reforms by pivotal actors within the legislative process.” While conceding the significance of transaction costs, I suggest that constituency costs have stronger effects. I will define constituency costs as voters' punishment of elected officials at the ballot box if a member of a political party defects from its ideological positions.

In arguing for the importance of constituency costs I assume that political parties and politicians want to stay in power and win elections, an assumption empirical evidence validates (Basinger and Hallerberg 2004; Kono 2006; Magee et al. 1989). Constituency costs, of course, depend on voters' attention and understanding of corporate tax policy. I argue that voters have limited information about policy changes depending on the level of policy complexity. The trade policy literature, which shows that complex policies enable officials to disguise their impact on voters, supports this argument (Guisinger 2009; Kono 2006; Magee et al. 1989). For example, optimal obfuscation theory posits that a party will choose more indirect policies for redistribution if the electoral gains in voter obfuscation exceed the electoral costs of receiving fewer resources from special interest groups (Magee et al. 1989). In a related argument, Kono (2006) argues that democracies increase non-transparent policy tools to optimize their support from voters and interest groups. Some trade barriers have more transparent effects than others. For instance, tariffs are import taxes, and voters tend to understand them, but non-tariff barriers and their effects are more complex. These differences in policy complexity have political consequences because politicians must weigh the costs and benefits of attacking status quo government policies. Thus, optimal obfuscation theory has clear implications for corporate tax policy transparency in that constituency costs can vary depending on the level of policy complexity. In turn, politicians may choose corporate tax policy instruments to minimize constituency costs and maximize their electoral benefits.

A broad range of policy instruments comprise corporate tax policy. Among them, I consider two main types of corporate tax policy: tax rates and tax incentives. The reason is that both policy instruments are mainly discussed in corporate tax reforms that are used to reduce the corporate tax rate and eliminate tax incentives with the goal of promoting business activities and maintaining stable revenue (OECD 2008). These two policy instruments have distinctive characteristics. At one extreme, tax rates are simply national tax rates. A change in the tax rate and its effects on tax revenues and their welfare are easily conveyed to voters. At the other extreme, tax incentives have complex effects whose impact on the welfare of voters is hard to explain. Since tax incentives take various forms, such as credit, deferral, and tax rate relief, and target specific industries or firms (Klemm 2010; OECD 2022), voters may find it hard to objectively evaluate how much taxes the firm is paying for its taxable income and its effects on their welfare. As a result, these two policy instruments differ from each other since tax incentives are more opaque policy instruments than tax rates. Thus, voters are far more likely to punish left-wing officials for deviating from their ideological position if they provide tax rate cuts than tax incentives.

The economic voting literature explains that voters tend to evaluate the economic performance of governing parties based on each party's traditional economic policies (Powell and Whitten 1993). Voters assess governing parties' performance based on the expectation that right-wing governments address inflation better than joblessness but left-wing governments address joblessness and income redistribution better than inflation. The left will face greater punishment at the polls than the right if they cut taxes, since voters expect the left to increase taxes to fund social expenditures (Hibbs 1977; Garrett 1998). There is anecdotal evidence of a partisan divide over corporate tax policy, particularly regarding corporate tax rates. In 2017, President Donald Trump signed the Tax Cuts and Jobs Act (TCJA), which introduced significant changes, including a reduction in the federal corporate tax rate from 35 percent to 21 percent. Democrats criticized the TCJA as a giveaway to large corporations and the wealthy, while Republicans argued that the tax cuts would stimulate the economy and enhance U.S. competitiveness (Sarlin 2017). Additionally, during the 2020 presidential campaign, Democratic candidates were asked about corporate tax rates. Buttigieg advocated for returning the rate to 35 percent, while Biden proposed increasing the federal corporate income tax rate to somewhere between 21% and 35%. These examples suggest that Democrats favored higher corporate tax rates, while Republicans supported lowering them. A poll also shows that Democrats favor increased taxes on corporations compared to Republicans (Pew Research Center 2017).⁴

Left-wing officials nonetheless have an incentive to lower corporations' ETRs. First, politicians regardless of party affiliation depend on corporate campaign contributions (Grossman and Helpman 1996). Firms influence public policy by a variety of means, and with a variety of objectives, but most corporations seek to lower their ETR (Richter et al. 2009). In the United States in particular, campaign spending has climbed significantly in recent years, increasing this dependence (Raja 2014).

Another incentive that left-wing politicians experience relates to constituent expectations. Conventional wisdom holds that left-wing politicians will lower unemployment rates (Hibbs 1977). A rise in corporate taxation, however, reduces employment by shrinking firms' new investment (Quinn and Shapiro 1991b). Thus, left-wing politicians keep corporate tax rates high to avoid constituent costs while at the same time, they need to promote economic growth by offering selective tax incentives to firms.

Right-wing parties, however, receive considerable rewards at the ballot box if they cut tax rates and far fewer constituency costs. Voters see their members' granting of tax cuts as a signal of good performance consistent with their traditional economic role. Thus, they can sell transparent tax cuts as a measure to grow the economy. My argument suggests that left-wing governments will be more likely to keep tax rates high but to provide more opaque policy tools than right-wing governments because of constituency costs.

Do partisan governments differ in collecting tax revenues from firms on absolute or relative terms? Quinn and Shapiro (1991a) find that Democratic administrations in the United States collect more taxes than Republican administrations. This may suggest that by offering limited tax incentives, Democratic administrations generate stable tax revenues. It could also imply something about the manner in which the left picks tax incentives to maximize tax revenue. Pinto and Pinto (2008) and Smith (2013) show that left-wing governments favor those firms and industries that generate the most employment for their core constituency, labor. According to Pinto and Pinto (2008), these governments therefore encourage the inflow of the type of investment that complements labor in production, while pro-capital governments promote the entry of investment that substitutes for labor. Smith (2013) also finds that left-wing governments provide bailouts to firms in labor-intensive industries. Indeed, the single biggest corporate tax breaks in the United States are for depreciation and expenses of capital equipment and production on US soil.⁵ The greatest beneficiaries of these breaks are the manufacturers who supported the Obama administration.⁶

To sum up, I propose that constituency costs affect political parties' incentive to enact opaque policies. Since voters easily evaluate the effect of tax rates on their welfare in contrast to tax incentives, political parties have more incentives to use opaque policy instruments like tax incentives to maximize electoral benefits. Yet such an effect is stronger for left-wing governments because they must balance the need for campaign donations from firms and economic growth by reducing their tax burden through tax incentives with low constituency costs and the need for voter support by avoiding policy instruments with high constituency costs. This discussion leads to the following hypothesis.

Hypothesis: Left-wing governments will be more likely to maintain high tax rates but compensate for high tax rates with generous tax incentives than right-wing governments.

I test my argument in the context of the United States for two reasons. First, the United States is a dominant actor in the global economy, constituting around 40 percent of all foreign direct investment inflows and 50 percent of all portfolio capital investment inflows (Swank 2013). Given this dominant role, US policies can strongly influence other countries' economic policies, including tax reforms. Thus, patterns that emerge in the voting behavior of US legislators may appear in the voting behavior of legislators elsewhere. Second, the standard view of the US corporate tax system is that its tax incentives are designed to stimulate investment and growth (Martin 1991). According to this view, political parties of any leaning will seek to promote investment and growth by cutting tax rates. Thus, the United States can serve as a hard case for testing the argument that partisan orientation affects corporate tax policy transparency. If I find partisan effects, other countries may reflect the same effects more strongly.

The data for this study consists of congressional roll call votes on corporate tax bills from 1986 to 2012 in the United States. Voting demonstrates legislative activity, showing how economic and political considerations affect legislators' voting behavior and how legislators take positions and claim credit on policy issues. I measure ideal points of legislators on roll call votes covering tax rates and tax incentives. I also examine what determines legislators' voting behavior on corporate tax policy transparency and report the results.

Data and Methods

I analyze the voting patterns of senators from 1986 to 2012 to examine how legislators vote on corporate tax policy concerning both rates and tax incentives. I also analyze the House votes to perform robustness checks for the results of the Senate votes and report them in the robustness check section. There are two main reasons I focus on senators' votes. First, the Senate roll-call votes show less partisanship than the House votes since senators answer to a larger constituency. Second, the Senate is less vulnerable to agenda control by the majority party since senators can offer amendments with fewer restrictions, possibly reducing selection bias.⁷ The period starts in 1986 because Congress implemented major corporate tax reform that year. Thus, looking at the changes after the 1986 tax reform allows us to examine the dynamics of partisan politics on tax bills.

I collected votes that determine the corporate tax rate and those that determine tax incentives in Vote View, relying on two criteria.⁸ First, the effects of the votes on these two areas of policy were clear. This includes the final passage of corporate tax bills that reflect a clear measure of legislators' preferences on corporate tax policy that reduces capital tax rates or increases tax incentives. Second, they would not affect other policy areas. All the selected bills received roll call votes in their respective chamber.

The criteria apply to five tax bills that affect tax rates in the study period: “Taxpayer Relief Act of 1997,” “Jobs and Growth Tax Relief Reconciliation Act of 2003,” “Tax Increase Prevention and Reconciliation Act of 2005,” “Tax Relief Unemployment Insurance Reauthorization and Job Creation Act of 2010,” and “The American Taxpayer Relief Act of 2012.” All were major legislative acts as reported by the Tax Policy Center of the Brookings Institution (The Tax Policy Center, 2015).⁹ Detailed information about tax rate bills is provided in Table A2 in the appendix.

The second set of votes provides tax incentives to certain businesses, which the bills call tax expenditures. I focus on the depreciation tax incentive among various types of tax incentives for two reasons. First, the depreciation tax incentive is one of the largest tax incentives provided in the United States. The report of the Joint Committee on Taxation states that the largest corporate tax expenditure is the domestic production activities deduction, with an estimated revenue loss of \$12.2 billion, and the deferral of income earned abroad with an estimated revenue loss of \$83.4 billion.¹⁰ Second, Congress extended the depreciation system in 1986, which significantly accelerated depreciation on tangible property, while repealing the investment tax credit (Sullivan 2011).¹¹ Thus, depreciation allowance is a unique case that will illuminate the effect of partisan politics on tax incentives. The report released by the Joint Committee on Taxation (2011) provides a listing of three tax bills that affected the depreciation system in the period of focus: “The Job Creation and Worker Assistance Act of 2002,” “The Economic Stimulus Act of 2008,” and “The American Recovery and Reinvestment Act of 2009.” Detailed information about tax incentive bills is provided in Table A3 in the appendix.

Each vote is coded as 1 if a senator casts a “yea” vote and 0 for a “nay” vote. Each voting legislator is coded as 1 for Republican affiliation and 0 for Democrat affiliation.

Statistical Methods

Data are the tax legislation on tax rates and tax incentives, which constitute binary responses to multiple votes from senators. To measure the ideal points of legislators on corporate tax bills, I employ an IRT model as a measurement model for binary responses.¹² IRT models link binary response variables to a single latent variable, showing the relationship between item responses and the latent variable. Item responses as observable indicators can help measure the latent variables indirectly.

I employ a measurement model because IRT models can provide vote-specific characteristics, such as difficulty and discrimination parameters.¹³ The difficulty parameter indicates the probability of answering an item correctly given the ability level of a respondent, suggesting the ease or difficulty of an item. The discrimination parameter represents how well an item discriminates among different trait levels. It indicates the strength of the relationship between the item and the latent trait, similar to loading in a factor analysis. Items with high discrimination parameters differentiate respondents more clearly (DeMars 2010).

Given these two parameters, IRT models allow the analysis to determine which items (here, votes) best predict the trait of the legislator (here, party affiliation). Another benefit of an IRT model is that a researcher can use it to evaluate or estimate the ability level of all legislators on the same continuum, as legislators might have different difficulties across bills. Besides the advantages of the IRT model, Bayesian inference also offers an advantage as it is appropriate to estimate the latent traits of legislators on multiple votes since it allows for estimating the uncertainty of parameters and treating unobservable parameters probabilistically.¹⁴

In understanding the structure of data based on the IRT model, the binary response is whether or not legislators voted “yea.” Legislators' latent trait is their underlying point of voting behavior on multiple votes. An item's difficulty indicates the degree to which votes relate to the occurrence of a voting behavior variable. An item's discrimination refers to the degree to which the vote distinguishes levels of the latent voting behavior variable. In the IRT model, the probability that a legislator i votes “yea” on a vote j can be written as:

Pryij=1θi,aj,bj=Φajθi−bjθi∼NXiβ,σ2

Where β, σ², and X_i represent a vector of coefficients, the variance of ideal points, and a vector of covariates for legislator i, respectively. θ_i is the ideal point of legislator i, a_j is a discrimination parameter for vote j, and b_j is a difficulty parameter for vote j.

To test my hypothesis, I conducted two analyses: measuring senators' ideal points on tax legislation and regressing their political party affiliation on these ideal points measured. The results from the statistical models support the hypothesis. The empirical analysis shows that Democrats tend to oppose reductions in tax rates while favoring tax incentives. Additionally, a regression analysis indicates that corporate tax policy is influenced by senators' political party affiliation. Overall, there is a strong relationship between senators' partisan orientation and the transparency of corporate tax policy.

Measuring Ideal Points of Legislators

Figure 1 presents caterpillar plots ranking Senators by their value of ideal points in both tax rates and tax incentives. As Figure 1 shows, parties have clear ideological differences between parties. Republicans have higher scores than most Democrats on tax rates in Figure 1(a), but they have smaller scores on tax incentives in Figure 1(b). These results suggest that Democrats tend to vote against reductions in tax rates but agree to vote for generous tax incentives, as expected. These findings are consistent with existing research showing that the Democratic party has a higher corporate tax rate and tax revenues in the United States (Inclan et al. 2001; Quinn and Shapiro 1991a; Quinn 1997). These findings also imply that Democrats oppose unpopular tax cut issues for their core constituency but court support from firms to increase their chances of winning elections. However, there is a lot of uncertainty about these estimates, which the width of most credible intervals reflects, since we have few items to estimate the latent factors.

In addition to the ideal points of legislators, Figure 2(a) displays the density plots of ideal points for each party and shows similar results: Republicans have higher scores than Democrats, and the distance between the median legislator of each party is wide in Figure 2(a). Furthermore, the location of the median legislator is between both parties, suggesting that senators from both parties are more polarized on tax votes. By contrast, the result of tax incentives shows the opposite: Democrats have higher scores than Republicans, and the distance between the median legislator of each party is huge in Figure 2(b). In particular, the location of the median legislator is closer to the location of the Democratic median, implying that Democrats have controlled the tax issue over time.¹⁵

Regression analysis of the effect of senators’ political party affiliation on tax legislation

In addition to the main findings, I conducted an additional analysis to examine the effect of senators' political party affiliation on corporate tax bills. The primary dependent variable is each legislator's ideal point, which the IRT model discussed in the main results above provides. My independent variable is a legislator's political party affiliation. I use a dichotomous variable, which measures whether a legislator is a Republican or Democrat. A legislator is coded 1 for Republican and 0 for Democrat.

To assess other factors affecting legislative voting behavior such as constituency interests, I include labor and exporting and importing industry interests in the models as shown in the previous studies that explore legislative behavior through roll-call votes on a wide range of foreign economic policies in the US context.¹⁶ Existing studies have shown that two types of constituent influences affect legislators' behavior: (a) class-based coalitions and (b) industry-based coalitions (Hiscox 2002; Jeong 2009). For example, trade policy literature demonstrates that the Democratic Party tends to represent the interests of labor and import-competing industries and favor trade protection. However, the Republican Party tends to represent the interests of capital and export-competing industries and favor free trade policy. This logic can be applied to corporate tax policy. Following Hiscox (2002) and Jeong (2009), I measure the interest of labor as total employment in manufacturing as a share of each statés total population. The dataset that I used was collected from the 1992 Census of Manufacturers and Census of Mining for the period between 1986 and 2010. The second dataset I use consists of measures of exporting industries and import-competing industries in each state. The measures of the weight of exporting and import-competing industries in a state reflect the total production of the top ten leading exporting and import-competing industries as a proportion of state income. The industry-level import and export data were from the U.S. International Trade Commission website based on the year 2002. The state-level industry production data were from the 1992 and 2002 Census of Manufacturers, Census of Agriculture, and Census of Mining. These variables are standardized so that the coefficients are comparable. I use datasets of constituents' economic interests that Jeong (2009) assembled.

Using data Prillaman and Meier (2014) assembled, I also include a set of other control variables, which cover employment rates (change in percentage of workers employed), growth rates of the gross product of the state, government ideology of the state, and union density in the manufacturing industry at the state level. These variables control for the confounding effects on the partisan orientation of legislators and voting on tax bills. Although these indicators capture the confounding effects, the causal arrows among predictors are ambiguous. Thus, including these control variables in the model can generate post-treatment bias that should not be included as control variables (Gelman and Hill 2006). For this reason, I report the results of models that exclude these variables in the results section.

I employ a linear regression model to examine the impact of political party affiliation on the tax bills: tax rate and tax incentive. The first dependent variable is the tax rate, and the second dependent variable is the tax incentive. Descriptive statistics for each of the variables appear in Table A1 in the appendix.

Table 1 presents the results of the analysis on the effect of political party affiliation on corporate tax bills: Models 1 and 2 focus on tax rates, while Models 3 and 4 examine tax incentives. The findings are consistent with the main findings reported in the ideal point estimates: Republicans are more likely to support reductions in tax rates, as expected. By contrast, the results of examining tax incentives tell a different story. As Model 3 indicates, Republicans are less likely to support tax incentive bills.

When it comes to the impact of constituency interest on legislators’ voting behavior, senators who represent the interests of export and import industries are more likely to vote to reduce the tax rate. On the other hand, senators who represent the interest of labor are unlikely to vote to reduce the tax rate. These results are consistent with the trade protection literature that suggests Democrats are closely associated with the interests of labor. Regarding tax incentive bills, senators that represent the interests of labor have a higher probability of voting “yea” while senators that represent the interests of export and import industries have a higher likelihood of voting “nay”. These results suggest that the high proportion of labor induces legislators not only to vote for reductions in capital tax rates, but also to vote to provide tax incentives. In addition, I further examine whether there is an interactive effect between political parties and their constituents' economic interests: the party to which a senator belongs and the proportion of labor in each state. I found that there is little effect between them. Thus, we can conclude that senators' voting choices are primarily determined by their political party affiliation in all tax bills, with constituents' economic interests -labor- influencing their votes to some extent on tax rate bills.

Two factors can explain the strong association between Democrats and labor on tax rate bills, as shown in Model 1 in Table 1. First, Democrats have shown a strong interest in reducing unemployment rates (Hibbs 1977). The manufacturing industry has historically played an important role in the growth of output and employment. It once provided high-wage jobs in the manufacturing sector and supported such jobs elsewhere. Jobs in manufacturing are seen as middle-class jobs. Helper, Krueger, and Wial (2012) report that the manufacturing sector accounts for 12 percent of the US economy and about 11 percent of the private-sector workforce, but that the country lost 41 percent of its manufacturing jobs between 1979 and 2009. Nonetheless, Democrats were highly supportive of manufacturing throughout the period of focus. Second, the labor sector makes most of its donations to Democrats. The labor sector has increased campaign donations and lobbying efforts over time despite decreasing political power arising from the decline in union membership and changes in economic structures.¹⁷ While they only donated \$76 million in the 2008 election cycle, they donated \$141 million to campaigns and committees in the 2012 election cycle (Vendituoli 2013).

In sum, there are clear differences between political parties in tax bills related to tax rates and tax incentives. I find that Democrats are more likely to oppose reductions in tax rates, but favor generous tax incentives aimed at promoting the manufacturing industry. The finding also shows that legislators' political party affiliation and constituents' economic interests-labor-determine legislators' voting behavior on corporate tax bills. Consistent with the ideal points measured, Democrats and legislators representing the interests of labor are more likely to oppose reductions in tax rates but favor generous tax incentives for the manufacturing industry.

Robustness Checks

I performed two robustness checks, using a model of votes on each agreement and all agreements in both the Senate and the House through a logistic regression model reported in Section C in the appendix. First, Figure 3 displays estimates without control variables using the senators' votes, and Tables C1 and C2 in the appendix present all results with and without control variables. The results of votes on all agreements related to both tax rates and tax incentives in the Senate are consistent with the theoretical expectation: Republican senators tend to support reductions in taxes and oppose the provision of tax incentives. Yet some results on each agreement do not correspond to the direction of the argument. For example, Republican senators tend to oppose two tax rate-related agreements: the Tax Relief Unemployment Insurance Reauthorization and Job Creation Act of 2010 and the American Taxpayer Relief Act of 2012. Furthermore, Republican senators tend to agree with tax incentives as stipulated in the Job Creation and Worker Assistant Act of 2002 and the Economic Stimulus Act 2012.

As the second robustness check, I run the same models using the House votes. Figure 4 displays the results without control variables, and Tables C3 and C4 in the appendix present all results with and without control variables. The findings are not consistent with the theoretical argument that Republican legislators tend to oppose reductions in tax rates but support the provision of tax incentives more than Democrat legislators do. Yet some results reveal that Republicans tend to agree with reductions in tax rates. These are the Jobs and Growth Tax Relief Reconciliation Act of 2003 and the Tax Relief Unemployment Insurance Reauthorization and Job Creation Act of 2010. When it comes to bills related to tax incentives, Republican legislators tend to agree with tax incentives across all models.

In summary, the results from the logistic regression models for each bill and all bills in the Senate align partially with the ideal point estimates of senators discussed in the main results section. Yet the findings of the House votes are not consistent with those of the Senate votes. One possible explanation for this discrepancy might be the presence of more complex dynamics in the House, such as higher levels of stronger leadership control and other factors that complicate voting patterns. These dynamics can make it more challenging to quantify the influence of partisanship in the House.

In this article, I have argued that the leftist party in the United States, the Democrats, tend to support high corporate tax rates but offer more generous tax incentives than the rightist parties. I explain that constituency costs influence political parties' incentives to adopt more opaque policies. Since voters easily understand the effect of some policy instruments, like tax rates, on their personal welfare, changes in tax rates will face higher constituency costs than opaque policies. At the same time, campaign donations tend to hinge on ETR. As well, lowering ETR promotes job production and growth. Politicians therefore have an incentive to lower rates, and left-wing politicians are more likely to minimize constituency costs and maximize their electoral benefits. Since left-wing governments pay a higher constituency cost for lowering corporate tax rates, they have a much stronger incentive to employ opaque instruments than right-wing governments.

The tax incentives Democrats favor are particularly aimed at promoting the manufacturing industry. The finding also shows the importance of the Democratic party's alignment with labor. Consistent with the ideal points measured, Democrats and legislators representing the interests of labor are more likely to oppose reductions in tax rates but favor generous tax incentives for the manufacturing industry.

However, there are some limitations in interpreting the results of this study. The single case study of U.S. corporate tax policy may limit the external validity of the findings when applied to corporate tax policies in other countries. This limitation could be addressed in future research that examines the effect of governing parties on tax incentives, both across countries and within countries, once comparable data on tax incentives become available. Second, the results should be interpreted with caution, as congressional bills often address multiple issues beyond corporate tax policy. Analyzing congressional speeches and sponsored bills could provide further insight into the net effect of political party affiliation on corporate tax policy. Future research could explore such data.

This study can contribute to existing literature on the role of partisan politics in corporate tax policy. This research provides a more nuanced explanation by showing that constituency costs have the potential not just to impede the corporate tax reforms, but also to give an incentive to political parties to create opaque policy. It shows that political parties tend to choose policy instruments with low constituency costs when voters differ in evaluating the effect of policy instruments depending on the level of policy complexity.

Moreover, I present a supply-side model, in which constituency costs affect incentives of political parties to choose an opaque policy over a transparent corporate tax policy. This supply-side model differs from a demand-side model, in which firms and voters pressure political parties to make corporate tax policies. In addition, this study enhances the generality of the policy transparency argument by extending it beyond the trade policy literature. While I have demonstrated the importance of transparency in corporate tax policy, future researchers might apply the findings to other policies, such as FDI and capital market liberalization.