Influence of Accounting Choices of
Corporate Donors on American Elections
Congress Elections of 2004 and Accrual Accounting Usage of
Organisations with a history of Large Political Expenses
Student: Ilona Knollenburg
Student Number: 10148205
Date: June 22, 2015
Version: Final
Word Count: 10,458 words
Education: MSc Accountancy & Control, track Accountancy University: Amsterdam Business School
Faculty of Economics and Business, University of Amsterdam Supervisor: dhr. Dr. W.H.P. Janssen
1
Statement of Originality
This document is written by student Ilona Knollenburg, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2
Abstract
This study focuses on the influence of the accrual accounting decisions made by corporate donors on the probability that a Congress candidate wins or loses the Senate elections of 2004. Thereby, providing additional insights on the accounting reporting decisions made by corporate donors of electable candidates in America and the motives behind these decisions.
Furthermore, this study contributes to the current literature by providing additional evidence regarding the political cost theory. Thereby contributing to the current discussion about whether or not this accounting theory should be extended.
In order to investigate this, data from Compustat, and hand collected date from Opensecrets.org, and CNN.com have been merged. 33 Senate races characteristics from 2004 are investigated with a final sample of 216 Contributions, which represent a total sponsored amount of $ 5,413,004.
The results indicate that abnormal accruals of contributing organizations do not significantly influence the outcome of Senate races. However, if a candidate is up for re-election this negatively affects their chances to win the race, as compared to a new challenger.
Thus, based on this study, it seems unlikely that the political cost theory needs to be extended. However, it is good to have estimated that the usage of accrual techniques does not directly pertain organizations to gather influence in the American politics, since the interest of society at large is not endangered in this way.
A limitation to this study is, that the focus lied on the total abnormal accruals from organization together. Thus, the possibility of different effects for positive and negative abnormal accruals have not been taken into account.
Key words: accrual accounting, accounting theory, agency theory, abnormal accruals, political cost theory, political science, lobby expenditures, Senate elections, corporate donors, America
3
Preface and Acknowledgements
With special thanks to:
My Parents, Jordan Smith, Roushan Oemar,
My fellow internship companions at PwC
for support, motivation, and advice. Without them, this thesis would not have been appeared in its current form.
4
Table of Content
Table of Content ... 4
List of Tables and Abbreviations ... 5
1. Introduction ... 6
2. Literature Review and Hypothesis Development ... 9
2.1 Background Of Lobbying ... 9
2.2 From the Agency Theory to the Political Cost Theory, and Further ... 10
2.3 Hypothesis Development ... 12
3. Research Sample ... 14
3.1 Sample Selection ... 14
3.2 Measuring Accrual Usage ... 16
3.3 Final Research Sample ... 17
4. Empirical Model & Analysis ... 19
4.1 Empirical Model Development ... 19
4.2 Descriptive Statistics on Sample ... 22
4.3 Analysis & Results ... 23
4.3.1 Correlation Analysis ... 23
4.3.2 Regression Analysis ... 24
5. Conclusion & Discussion ... 28
Reference List ... 31
5
List of Tables and Abbreviations
List of Tables
Table 1 Panel a. Sample selection: Congressional races to be investigated ... 15 Panel b. Sample selection: Congressional races to be investigated (continued)
and PAC contributors ... 15 Table 2 Panel a. Sample selection: PAC contributors to be investigated (Continued).. 18 Panel b. Final sample figures ... 18
Table 3 Descriptive Statistics for Variables Influencing Election Outcomes ... 23 Table 4 Correlation Table for Main Effects (N = 214) ... 24 Table 5 Hierarchical Regression Analysis for Variables Influencing Election
Outcomes (N = 214) ... 25
List of Abbreviations
FEC Federal Election Commission
PAC Political Action Committee
SEC Securities and Exchange Commission
6
1. Introduction
Correia (2014) published in her article that from the 4008 investigated restated firm-years 19.26 percent included amounts of money spent by organizations on both direct and indirect lobbying. In this way, organizations are able to influence the American politics. Gaining influence in the decision making process regarding laws and regulation is usually negatively perceived by society, as it enables organizations to get laws and regulations formulated in their own best interest. This is a logical relationship, since the best interest of organisations usually differs from the interest of society. Thus, actions have been undertaken to make organizational political spending more visible (Lobbying Disclosure Act of 1995).
However, the main problem here is estimating how much influence organizations gain with their political expenditures, since it is not something you can easily measure, let alone express in figures. To provide more visibility in the power of the organizations and to prevent advocacy in the politics in America the lobby expenses made by companies are required to be recorded, and made publicly available by law since 1995 (Lobbying Disclosure Act of 1995). But, what if lobbying isn’t the only way in which organizations can or try to gain influence in America’s politics?
It is highly likely that there are a number of ways for organisations to get some influence in the decision making process regarding laws and regulations by governments. One way is through direct lobbying, as investigated Correia (2014). But influence can also be gained in an indirect way. However, unlike direct lobbying which can be estimated quite reliably due to reporting regulations on the subject (Lobby Disclosure Act of 1995), estimating the impact of and the ways in which organisations try to obtain influence through indirect lobbying is much harder to detect.
A possible way of indirect lobbying by organisations is applying politically motivated earnings management (Fields, Lys & Vincent, 2001). Ramanna & Roychowdhury’s study (2010) provided additional evidence for the assumption that corporations use accounting discretion during election years to manage political costs. Thus, implying that accounting discretion could be used by organizations to protect the supported candidate from scrutiny over the accounting figures of a hot election issue, and gain influence in politics by doing so.
Further evidence on this matter is provided by Kido, Petacchi & Weber (2012). They looked at gubernatorial elections and state government’s accounting choices. They investigated the compensated absence liability account and the unfunded pension liability
7 account, since there is flexibility for manipulation in these accounts (Kido, Petacchi & Weber, 2012). Their results suggest that state governments use different accounting measurements to present a healthier financial picture in an election year.
On the other hand Van Lent (2012) states that the association between elections and discretionary accrual measurement as was found by Kido, Petacchi & Weber (2012) may reflect other factors like the state’s dependence on credit markets. Due to the contradictory opinions, additional research is needed to clarify the relationship between the accrual usage of corporate donors and the real influence gained in politics by organisations. This has been indicated by Ramanna & Roychowdhury’s (2010) as well. With this study, I want to provide additional evidence on this topic. Which leads to the following research question: “What influence did the accrual choices of corporate sponsors have on the probability of Congress candidates to win the Congress election of 2004 in America?”.
This study contributes to the growing literature on the subject of the relation between elections and the accounting choices of (non-governmental) organizations. More specifically, this study studies the effect of the political cost hypothesis, and whether or not this hypothesis is potentially more evolved than currently assumed. Since the findings of the studies of Ramanna & Roychowdhury (2010) and Kido, Petacchi & Weber (2012) are contradictory to the findings of Van Lent (2012) additional research is needed on this subject to fill the gap in the current literature.
Also, by testing the influence of elections an accounting accruals important insight can be gained on how elections influence organizations accounting decisions. This is relevant from a societal point of view, because corporations try to influence the decision making process in their own best interest (Fisher & Eisenstaedt, 2004). However, these interests are unlikely to be the same as the interest of society in general. Furthermore, since corporations have indeed an additional way to influence the political environment, it is necessary that society becomes aware of this possibility in order to invent an adequate control measurement (like the Lobbying Disclosure Act 1995 does for lobbying expenses) and to ensure that politicians act in the best interest of society. In this way, these actions can be made visible, and advocacy can be prevented.
For this research, ddata has been gathered from Compustat, and this is combined with hand collected data from Opensecrets.org, and CNN.com on the Senate elections of 2004 to answer the research question. I investigate 33 Senate races from 2004, leading to an initial sample of 66 Congress candidates. After removing candidates from which the corporate sponsors had incomplete data available to estimate their accrual accounting usage with the
8 Jones model (1991), the final sample consists of 50 Senate candidates that are sponsored for a total of $ 5,413,004 by 214 contributions received from corporate sponsors. Lastly, a regression to estimate the probability of a candidate to win or lose the election based on the total abnormal accrual usage of their corporate donors is formulated.
After the empirical model has been ran on the gathered data, two outcomes where of main importance. The first finding indicates that the abnormal accruals of contributing organizations do not significantly influence the outcome of Senate races on their own, since the effect of the abnormal accruals is largely insignificant (p-value being 0.291). Thus, it is unlikely that the political cost theory should be extended, as was argued by Ramanna & Roychowdhury (2010). Also, it is good to have pointed out that the abnormal accrual usage of large sponsoring firms does not provide this firms with an indirect lobby opportunity.
The second outcome is related to the control variable incumbent. A positive effect of a candidate being incumbent on the probability on the election outcome for a candidate was expected based on prior literature. However, my findings indicate that a candidate being incumbent negatively affects the possibilities of that candidate to win the elections. This could provide an opportunity for further research, as no explanation has been found for this uncommon outcome.
The remainder of this paper is organized as follows. First, there will be information provided regarding the theoretical background of this subject, and the hypothesis will be formulated as well. Secondly, a description of the methodology and the sample selection is provided in section 3. This is then followed up by section 4, which first develops the empirical model, and a discussion of the results after running the empirical model. Lastly, a conclusion and suggestions for further research, as well as the limitations of this study are presented in section 5.
9
2. Literature Review and Hypothesis Development
Section 1 provided some information on the background of this study. This section discusses the literature background in further depth by first providing an overview of the theoretical concepts behind lobbying, which is followed by an overview of the agency theory, and the development from the agency theory to the political cost theory. Also, the connection between the agency theory and the use of accruals by contributors will be pointed out. Lastly, the hypothesis for this study is formulated based on the current literature and theoretical constructs.
2.1 Background Of Lobbying
As was stated in the introductory paragraph, 19.26 percent of the S&P 500 companies investigated by Correia (2014) participate in either direct or indirect lobby activities. From this can be concluded that quite a large amount of companies that already have quite an influence on society due to their firm size, also try to extent their influence in the American decision making process.
Based on the agency theory, a rational individual only participates in lobby activities if he or she expects future benefits that exceed the cost from undertaking these activities (Sutton, 1984). Lobbying has been defined by Sutton (1984) as follows:
“The efforts of individuals and organizations to promote or obstruct new regulations, whatever the source, are described collectively as lobbying.”
One way for organisations to enact in accordance with the Lobby Disclosure Act of 1995 is to register by the Political Action Committee (PAC) (Ramanna & Roychowdhury, 2010). The data on the so called PAC contributions is publicly available at Compustat and/or the Federal Election Commission’s website (www.fec.gov). In this way, the Lobby Disclosure Act (1995) tries to make political contributions and connections more visible, resulting in a more transparent decision making process for society. Transparency is enhanced, by showing the political contributions made by organisations and to whom they are made. Since this information is publicly available, it should prohibit governmental decision makers from acting only in the best interest of the lobby organisations (Fisher & Eisenstaedt, 2004). Thus, this should result in actions of politicians that reflect the best interest of society in general.
10 Although the Lobby Disclosure Act has been included in the American Law for several years, companies still participate in lobby activities. If companies participate in political expenditures to gain direct influence in the politics because they expect benefits resulting from these expenditures, it is likely that they are also trying to gain influence in indirect ways (Correia, 2014).
From the accounting literature, the desire of organisations to participate in lobby activities is mainly explained from an economical point of view. Watts (1974; 1977) pointed out that it is reasonable to assume that managers act in their own best interest. So, if management has concerns about negative publicity due to their reported figures, firms are likely to either alter their financial reporting figures by using earnings management tricks (Guay, 2010; Jones, 1991) or to pressure their political connections in order to obtain more “suitable” reporting regulations (Sutton, 1984). This idea of politically motivated earnings management based on agency-related incentives (Fields, Lys &Vincent, 2001), and the political cost theory will be discussed more in depth in subparagraph 2.2.
Apart from the economical reasoning, there are also arguments arising from the political science literature to explain the desire of organisations to lobby. Alexander (1989) argues that money is an element of political power, since it buys what cannot otherwise be obtained. This can be the case for both the politician and the firm manager, since the politician can use the money for their promotion campaign, and the firm managers can use money for lobby activities. Thus, resulting in political power (Alexander, 1989).
Moreover, since the money available to a politician and its party affects political spending, money is of fundamental importance for politicians (Johnston, 1985). Therefore, a logically opportunity is created for organizations to gain influence in politics by making political expenditures. This statement has been proven by Hart (2001), Snyder (1992), and Kroszner & Stratmann (2005), whom concluded that political candidates and their donors exchange favors. And in this case, the favors being influence for the organizations, and money for the political parties.
2.2 From the Agency Theory to the Political Cost Theory, and Further
Watts (1974; 1977) started with his paper the discussion of the form of the financial statements, and more specifically wanted to explain why the financial statements are presented in their current form. He suspected that the rules for the financial statements are a product of both the market and political forces. But, since it can be assumed that individuals act in their own best interest according to the agency theory, it would be logical that corporate
11 management tries to influence the political procedures based on their attitudes towards the accounting standards (Watts, 1977; Watts & Zimmerman, 1978).
This has led Watts & Zimmerman (1978) to develop a positive theory of accounting, the so called political cost theory. The political cost theory states that larger firms are more visible, and more likely to attract public attention when declaring “high” profits, thus more likely to choose accounting procedures that defer reported earnings from current to future periods to avoid high taxes and to avoid drawing the public’s attention on the company (Watts & Zimmerman, 1978; Ramanna & Roychowdhury, 2010). Organisations are likely to incur political costs such as union expenses and regulatory oversight boards, because society tends to view “high” reported earnings of large companies as unreasonable and as signs of a monopoly position, thus creating an intention for the government to undertake adverse political actions. Likewise, the thread of these political costs gives incentives to management to lobby on accounting standards, since the reported profit of the corporation depends on these accounting standards. So, the avoidance of political costs and the public scrutiny is partly based on corporate managements own self-interest (Watts & Zimmerman, 1978).
Since the introduction of the political cost theory more studies have been undertaken to explain the behaviour of companies regarding their financial disclosures in light of this theory. Sutton (1984) also pointed out how the relationship between management of organisations is affected by changes in financial accounting standards and the actions that are undertaken by management to get these standards adjusted in their advantage. These actions are likely to be lobbying, but in many creative and disguised ways (Sutton, 1984).
Where Sutton’s research is mostly dedicated towards lobby actions, Jones (1991) focused his research on income decreasing earnings management to get favourable import relief determinations from the United States International Trade Commission (USITC). He estimated that companies which are subjective to import reliefs report lower earnings since this is one of the indicators for the USITC (to subject an organisation to import reliefs). The USITC investigates requests for import reliefs up to 5 years prior to a petition that has to be filed by the organisation. The commission looks for a downward trend in financial performance or a drastic decline in recent periods (Jones, 1991). This results in an expectation of income decreasing accruals in general for organisations that perform import activities, since they are likely to file for an import relief.
Also the study of Correia shows that organizations are able to influence independent agency bodies, as well as politicians, through political expenditures. The findings of the study indicate that organisations that contribute to politicians in a strong position are able to put
12 pressure on the Securities and Exchange Commission (SEC), which results in a lower likelihood of being prosecuted by the SEC (2014).
In all of the previously mentioned researches management undertakes activities that directly or indirectly involve accounting measurements that are used to prevent political costs. However, the research of Ramanna & Roychowdhury (2010) created some uncertainty surrounding the completeness of the political cost theory. They argue that, based on their research results, it is also possible that organisations gain influence in the American Politics by the use of income increasing or decreasing accruals, rather than just avoiding negative publicity with them. This is based on their finding that firms with a history of campaign contributions to Congress candidates have incentives to take income-decreasing accounting actions in 2004 compared to a prior election year if they engage in outsourcing (Ramanna & Roychowdhury, 2010). The reason behind this appears to not only be the avoidance of political cost, but also to protect the candidates which they contributed too from scrutiny, making it more likely that these candidates will be chosen by the voters.
On the one hand it seems that Kido, Petacchi & Weber’s (2012) research support Ramanna & Roychowdhury’s study. Their findings indicate that the state government’s accounting choices were adjusted to show a better financial situation in an election year for the compensated absence liability account and the unfunded pension liability account.
On the other hand, Van Lent (2012) disagrees with the findings of Kido, Petacchi & Weber (2012) based on his study on incentives of governor’s for changing accounting measurements in years of elections. He states that the existence of disciplinary mechanism and the incentives to be transparent derived from voter interest, capital markets, the media and political competition weight heavier that the private benefits that politicians might perceive from accounting manipulation. Moreover, states do not necessarily have to use accounting tricks, because they are allowed to choose their own accounting policy (Van Lent, 2012). Thus, the effect of elections on incentives for the use of accounting accruals in governmental organizations can not as naturally be assumed as Kido, Petacchi & Weber (2012) did (Van Lent, 2012).
2.3 Hypothesis Development
Now that both the economical and political science theoretical background for this study have been developed, it is time to formulate the hypothesis. With this research, I want to look further into the relatedness of the use of income increasing and decreasing accruals by corporate donors as an indirect lobby possibility. The motive behind this is to find out if
13 companies indeed gain influence in the decision making process by using accruals in certain ways around election years. If this is the case, then there is new evidence gathered for the extension of the political cost theory as argued by Ramanna & Roychowdhury (2010).
Elections are expected to provide incentives for firm managers to use accrual based earnings management if these firms participate in long-term political sponsoring of candidates. The organizational disclosures are expected to be adjusted downwards by firm managers if there are concerns about negative publicity and political scrutiny (Guay, 2010; Ramanna & Roychowdhury, 2010). Thus, it can be expected that election years provide additional incentives for downward adjustments as compared to non-election years, since elections enhances the attention to the financial disclosures of supporting firms (Fields, Lys & Vincent, 2001).
Adjustments of financial disclosures by management is called earnings management, which can be divided in real earnings management and accrual earnings management (Chi, Lisic & Pevzner, 2011). Only accrual based earnings management will be included in this research, since these can more easily and reliable be identified than real earnings management (Jones, 1991). Since the elections provide incentives to report more “suitable” disclosures, the agency theory can be used to explain the incentives for managers to use accrual based earnings management in years of election. And because it are mostly the earnings figures of the contributors that are taken into account when evaluating firms and their motives (Hart, 2001), the agency theory and accrual based earnings management are combined in this study.
In short can be stated that recent research has not solved the discussion surrounding the reasons for why organisations who have a history of political contributions to use income decreasing accruals in years of elections. Although not all the literature that has been discussed is equally convincing, there is reason to believe that the political cost theory as defined by Watts & Zimmerman (1978) does not capture the complete story behind the motives of management to gain influence in the decision making process of standard setting. Therefore, I think that it is likely for organisations that participate in earnings management in years of elections to use income decreasing accruals in order to not only avoid political costs, but also to enhance the chance from their supported candidates to win the elections.
This leads to the hypothesis to be formulated as follows:
Hypothesis 1: Electable candidates have a greater (weaker) chance to win an election if a corporate sponsor uses more income decreasing (increasing) accruals in election years.
14
3. Research Sample
Section 2 provided an overview of the current literature regarding the political cost theory, and the current discussion surrounding this theoretical concept. Also, the hypotheses were formulated. This chapter provides detailed information about the sample that will be used to answer the research question. Also, more information is provided about the data sources and the eliminations that have been made in order to select candidates and corporate donors. This is followed by explaining the measurement for identifying accrual earnings management used.
3.1 Sample Selection
For this research, data has been looked up for the Senate elections of 2004. The Senate elections of 2004 have been chosen specifically to limit the scope of this research, and to provide evidence for the extension of the political cost theory, as assumed by Ramanna & Roychowdhury (2010). Choosing the same election year is likely to mitigate biases that could occur due to different factors surrounding the elections in different years.
Also, it is interesting to look at the Senate elections specifically, because of the influence Senate members have in the American Congress. Senate members are known for their ability to guide their parties and their power to influence the Senate, and thus the Congress, in determination of a legislative form (Senate.gov, n.d.). The influence of the Senate provides incentives for large organizations to point their lobby activities directly to Senate members and Senate candidates. Furthermore, the Senate is chosen for a longer time, and therefore better able to act independently of the public opinion, and thus more likely to act on their own self-interest (Senate.gov, n.d.).
After the elections of 2004 the 109th Congress will be formed (OpenSecrets.org). The
Senate consists of 100 members: 2 members for each State; the District of Columbia is excluded from the right to appoint a senator by law. Each Senate member serves for 6 years, and elections take place in even years. Meaning that for each election round, about one third of the Senate is up for (re)election (Senate.gov, n.d.; CNN, 2005). For this study, all Senate races of 2004 are identified by using the sources OpenSecret.org and CNN (2005), Following Ramanna & Roychowdhury (2010), I look at the Democratic and Republican candidates per race only, thus excluding third-party candidates, because of the limited percentage of votes they get compared to Republican and Democrats candidates. Furthermore, races in Louisiana are excluded since multiple candidates from the same party can contest in elections there.
15
Table 1
Panel a. Sample selection: Congressional races to be investigated
Panel b. Sample selection: Congressional races to be investigated (continued) and PAC
contributors
Panel a. Start Sample Selection
Races Candidates
Initial Sample 34 136
-/- Excluded States (Louisiana) -/- 1 -/- 7 -/- Non-Democratic/
Republican Candidates
- -/- 63
Subsample 1a. 33 66
Panel b. PAC Contributors
Races Candidates Contributors Contributions
Subsample 1a 33 66 1327 $ 35,603,229
-/- Non-PAC Sponsored Candidates
- -/- 1 - -
Subsample 1b 33 65 1327 $ 35,603,229
These criteria result in a subsample of 33 races and 66 candidates (see Table 1 Panel a., and CNN, 2005).
The next step is identifying donors that have long-term relationships with the Senate candidates, and are considered to be top contributors. In order to gain information about these relationships, the site OpenSecrets.org is used. This site contains data on PAC contributions retrieved from the Federal Election Commission’s (FEC) databases about Congressional races by state. Consistent with Ramanna & Roychowdhury (2010), I select only 20 largest (top) contributors per candidate in the election and pre-election period 2003 - 2004. By looking only at the largest contributors, corporate donors can be selected which made large contributions to at least one candidate. This selection is based upon prior literature that has found evidence about large (cash) donations being used to establish links between politicians and firms (Kroszner & Stratmann, 2005; Hart, 2001).
After applying these criteria, the sample of Congress races that are going to be investigated is 33. However, candidate Sowell from the jurisdiction Alabama only received private sponsoring. Thus, candidate Sowell is excluded from the initial number of candidates to be investigated for this research. The initial number was 66 candidates, so this results in 65
16 candidates. They received money from 1327 contributors which contributed a total of $ 35,603,229 (as counted by top contributors per candidate, no reduction has been made for organisations that sponsor more than one candidate), see Table 1 Panel b.
3.2 Measuring Accrual Usage
As has been discussed in chapter 2, elections provide an incentive for managers of organizations with a long-term political sponsoring history to adjust the organizational disclosures downward if there are concerns about negative publicity (Guay, 2010; Ramanna & Roychowdhury, 2010). Adjustments of financial disclosures by management is called earnings management, which can be divided in real earnings management and accrual earnings management (Chi, Lisic & Pevzner, 2011).
Real earnings management occurs when earnings are managed by manipulating real activities, for example temporary price discounts. These actions have negative consequences on future cash flows and negative effects on the firm value in the long run. However, real earnings management is unlikely to draw attention, as long as the actions are properly disclosed in the financial statements (Chi, Lisic & Pevzner, 2011).
Accrual earnings management is applied when accounting techniques are used to produce financial reports that may paint an overly positive picture of a company (Chi, Lisic & Pevzner, 2011). Since accounting techniques are used, investors might be fooled by the figures, but no other harmful actions for the organization are undertaken. Thus, accrual earnings management is less costly for a firm than real earnings management.
For this research, I focus only the use of accrual earnings management by organizations, since these can more easily and reliable be identified than real earnings management. To identify real earnings management, subsequent measures need to be established. Accrual based earnings management can be calculated with the Jones Model, and the financial statement of the organization (Jones, 1991).
Total accruals and components of accruals are regularly used in studies to measure earnings management. Especially, total accruals are interesting to identify earnings management, since managers are likely to use several accruals to alter reported earnings (Jones, 1991). However, accruals related to income taxes are excluded, since these can affect multiple years and the focus of this research lies on comparing the accruals of 2003 and 2004 with each other. Thus, this research focus on the total accruals of organizations’ earnings before taxes to capture a large portion of accrual based manipulations.
17 𝑇𝐴𝑡= [∆ 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑡 (𝐴𝐶𝑇) − ∆ 𝐶𝑎𝑠ℎ𝑡(𝐶𝐻𝐸)]
− [ ∆ 𝐶𝑢𝑟𝑒𝑛𝑡 𝐿𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠𝑡(𝐿𝐶𝑇) − ∆ 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑖𝑒𝑠 𝑜𝑓 𝐿𝑜𝑛𝑔 − 𝑇𝑒𝑟𝑚 𝐷𝑒𝑏𝑡𝑡(𝐷𝐷1) − ∆ 𝐼𝑛𝑐𝑜𝑚𝑒 𝑇𝑎𝑥𝑒𝑠 𝑃𝑎𝑦𝑎𝑏𝑙𝑒𝑡(𝑇𝑋𝑇)] − 𝐷𝑒𝑝𝑟𝑒𝑐𝑖𝑎𝑡𝑖𝑜𝑛 𝑎𝑛𝑑 𝐴𝑚𝑜𝑟𝑡𝑖𝑧𝑎𝑡𝑖𝑜𝑛 𝐸𝑥𝑝𝑒𝑛𝑠𝑒𝑡(𝐷𝑃𝐶)
Where the ∆ is computed between time t and t-1 (Jones, 1991), and company firm data is retrieved from Compustat. Based on the expectations model from DeAngelo (1986), the use of total accruals from the prior period (t-k) can be used as a measure of ‘normal’ accruals. ‘Abnormal’ accruals can then be determined as the difference between the total accruals for the current period and the identified level of normal total accruals.
So, the level of total accruals used in 2003 can be seen as the normal accruals level and the difference between the levels of total accruals used in 2004 minus the accruals used in 2003 as the abnormal accruals (Jones, 1991; Figure 1 – Timeline for measuring abnormal accruals). This assumption is funded on the political cost theory, i.e. managers are expected to have additional incentives for the use of accruals on the financial disclosures in years of elections (Fields, Lys & Vincent, 2001), and political science literature which shows that political candidates and their donors exchange favors (Snyder, 1992; Krosner & Stratmann, 2005; Hart, 2001).
Total Accruals Prior Period Total Accruals Current Period
2002 “Normal Level Of Accruals” 2003 Election Period 2004
Figure 1 – Timeline for measuring abnormal accruals. Abnormal accruals are calculated as the value of the total accruals of the current period -/- the total accruals of the prior period (i.e. normal level of accruals).
3.3 Final Research Sample
Having introduced the concepts of ‘Normal Level Of Accruals’ and the ‘Abnormal Accruals of the Current period’, the final research sample can be identified. The initial sponsor list consisted of 1327 contributors for 65 candidates. From this initial sample, some of these contributors should be excluded. One kind of contributors that is excluded from the sample, are financial institutions. These contributors are excluded due to their specific characteristics. Also, contributions are excluded due to missing data needed for establishing accrual usage by the sponsoring firms.
18
Table 2
Panel a. Sample selection: PAC contributors to be investigated (Continued) Panel b. Final sample figures
Panel a. PAC Contributors
Races Candidates Contributors Contributions
Subsample 1B 33 65 1327 $ 35,603,229
-/- Contributors With Missing Values on Accrual Data
- -/- 15 -/- 1111 -/-
$ 30.190.225
Final Sample: 33 50 216 $ 5,413,004
Panel b. Final Sample Figures
Winning Candidates Losing Candidates Democrats Republicans Initial Sample: 33 32 32 33
-/- Candidates with Missing Values of Contributors
- -/- 15 -/- 10 -/- 5
Final Sample: 33 17 22 28
In total are 1111 from these 1327 contributors excluded, leading to a total of 216 contributions from 131 firms with long-term relationships with Senate candidates for the elections of 2004. These 216 donors contribute a total of $ 5,413,004 to the 50 Senate
candidates of 2004, see Table 2). Note that these exclusions only resulted from missing data on contributions from firms to candidates, missing data on accruals for contributing firms, and firms who have specific characteristics due to their industry segment.
For this final sample, an empirical model will be developed in chapter 4. With this empirical model I will be able to look at the effect of the abnormal accruals used by the firm who made a contribution on the election outcome individually. Thus, N = 216 given that this number still has to be corrected for outliers, which will also happen in chapter 4.
19
4. Empirical Model & Analysis
Section 3 used the hypotheses as a starting point for identifying the research sample, and determinants. Also the relationship between them has been pointed out. In section 4 the empirical analysis takes place. First a regression to answer the research question, i.e. the way in which the probability that a candidate wins or loses the election is affected by the accrual choices of their corporate donors, is developed. This is followed by a discussion of the findings, from which the conclusion will be drawn in section 5.
4.1 Empirical Model Development
This paper performs quantitative research, based on data from publicly available databases, and sources of firms in America with a history of sponsoring politicians. In order to perform quantitative research, a suitable empirical model is needed. Data about income accruals of corporate donors need to be compared with data on political expenditures and the probability of sponsored candidates to win or lose an election to form new information. Thus, this section is devoted to establishing the regression and explaining its variables.
Hypothesis 1 states that: Electable candidates have a greater (weaker) chance to win an election if a corporate sponsor activities use income decreasing (increasing) accruals.
In order to test hypothesis 1, I combined the accrual accounting data from Compustat with the data available on the Senate elections of 2004 from Opensecrets.org. Hereby, identifying 1327 Contributors that have both a large and long-time interest in these elections. After controlling for missing values, a final sample of 216 firms and 50 Senate candidates are identified for the elections of 2004.
To generate results which can lead to the acceptation or rejection of Hypothesis 1, a suitable measure and the components for the empirical model needs to be established. I want to look at the effect of each of the abnormal accruals of the contributors on the election outcome per candidate. Therefore, the candidate specific data will be taken into account per contribution made to that candidate. Thus, the initial number of N will be 216 contributions. The correction for outliers is applied in subparagraph 4.2.
The first component of the model is the dependent variable (y): the election outcome per candidate of the Senate election races of 2004. This is measured in the following way: win = 1; lose = 0.
20 The component that will be the central variable in the model is related to accrual usage by sponsors. The use of income decreasing accruals result in lower reported figures for organisations. Ramanna & Roychowdhury (2010) found that firms use more accruals in the election year of 2004 compared to previous years, and thus have lower reported earnings in years of elections. Assuming that firms indeed use more accruals in years of elections, and that the political cost theory indeed should be expanded, I take the amount of standardized abnormal accruals per organisation as main explanatory variable in the regression. Thus, the first variable of the empirical model is β1 Z Abnormal Accruals, where the “Z” represents the standardisation of the variable.
The abnormal accruals have been standardised in the regression, because the variable β1 Z Abnormal Accruals is measured by expressing its value in dollar amounts, whereas the outcome of the election is a win or lose. This could make it difficult to compare the abnormal accruals of a contributor with the election outcome of a sponsored candidate. To remove the differences of scale in the empirical the variable Abnormal Accruals is standardized in SPSS following the following formula:
𝑍𝑖,1𝜎 =𝑋𝑖 − 𝑋𝑆 𝜎𝑋,𝑆
Where 𝑍𝑖,1𝜎 = the data point i standardized, 𝑋𝑖 = each data point i,
𝑋𝑆 = the mean of the whole sample,
𝜎𝑋,𝑆 = standard deviation of the whole sample.
The expected effect of β1 Z Abnormal Accruals on the probability of a candidates to win the election is positive, because lower reported earnings figures give less possibility for public scrutiny over these figures, and thereby avoid negative news on their candidates. Thus, increasing a candidate’s probability of winning.
The second variable in the model is the control variable C2 Incumbent. Incumbent explains the effects on races of candidates that are up for re-election. Incumbent is an explanatory variable, because prior literature states that candidates who are up for re-election have a greater chance to win a Congressional race. Moreover, they mostly win with large margins too (Mayhew, 1974; Alford & Hibbing, 1981; Green & Krasno, 1988). This is the case, because candidates who are up for re-election have a number of advantages over challengers. The first advantage was indicated by Tufte (1973), and is related to the gerrymandered districts. Also, a high name recognition by voters is an advantage (Campbell,
21 Converse, Miller & Stokes, 1966). Lastly, Fiorina (1977) indicated that incumbent candidates have more opportunities to do favours for their contributors, which leads to the last advantage of Incumbent candidates: ability to raise and spend large sums of money on the re-election campaign (Green & Krasno, 1988). This also makes candidates who are up for re-election an interesting subject for lobby organizations
In short can be stated that, based on prior literature, a candidate who is incumbent has a greater chance to win an election. So the expected effect of C2 Incumbent on the probability of a candidate to win an election is positive. This effect is going to be measured with a dummy variable, being 1 if a candidate is up for re-election and 0 if the candidate is a new challenger.
Lastly, a second control variable is added to the empirical model: C2 Republican. This control variable is set of to control for the effects of the popularity of one party over the other in the election year due to other factors than the abnormal accruals of the sponsoring contributors or a candidate being Incumbent or not. In the election year of 2004, Americans were closely divided between the Republican party and the Democrats (Abramowitz & Stone, 2006). Therefore the party to which a candidate belongs is suitable to be a control variable, since the probability of a candidate to win the elections of 2004 are expected to be equally distributed (Sunshine-Hillygus & Shields, 2005). Accordingly, the expected effect of the control variable on the overall model is +/-. This control variable is also a dummy variable, being 1 if a candidate is Republican, and 0 if a candidate is a Democrat.
Now that the relevant variables are pointed out, the empirical model can be developed. By keeping in mind the probable effects of the variables on the probability of a candidate to win or lose an election, the following empirical model has been developed to estimate the probability of a candidate to win or lose the election based on the accounting choices of their largest contributors in election years:
𝑦 = 𝛽0 + 𝛽1 𝑍 𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝐴𝑐𝑐𝑟𝑢𝑎𝑙𝑠 + 𝐶1 𝐼𝑛𝑐𝑢𝑚𝑏𝑒𝑛𝑡 +/−𝐶2 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 + 𝜖
Where y = election outcome per candidate for the Senate election of 2004, β0 = a constant,
β1 = variable for the effects of the standardized abnormal accruals of the contributing firms per candidate,
C1 = control variable for the effects of the candidate being Incumbent (or not), C2 = control variable for effects of being a candidate for a party (Democrat or Republican only),
22 This model contains two explanatory variables, from which one is also a control variable, and one additional control variable. Initially, the idea was to also include a third control variable for the amount of contributions that have been contributed to the candidates and how the contributions have been divided over the candidates in a Congress race, since this was expected to function as an explanatory variable as well. This was expected, because the amount of contributed money enables candidates to promote themselves, which in turn result in voters to have opinions on candidates. This strongly influences how voters vote (Jacobson, 1978). Resulting from this, candidates who collect more contributions than their opponents have a higher change to win the elections.
Although Green & Krasno (1988) found that Incumbent candidates are better able to raise large amounts of contributions, when looking at the final sample it can be stated that there are not very large differences between the contributions received for challengers and incumbent candidates. Furthermore, Jacobson (1978) found that spending (and collecting) by challengers have a greater impact on the outcome of an Senate election that spending by candidates who are up for re-election. Thus, even if Incumbent candidates receive slightly more contributions in total, this should not have a significant influence on their chances of winning. Therefore, the amount of contributions is excluded from the empirical model.
4.2 Descriptive Statistics on Sample
This section provides information on the descriptive statistics on the sample, and also identifies outliers which could affect the results and analyses of this study.
Firstly, the mean and the median for the sample of 216 contributing firms are compared to identify possible outliers. It can be concluded that the mean, and the median differ significantly (see Appendix 1). This means that there are certain outliers in the sample. An inspection of the normal Q-Q plot of Abnormal Accruals (see Appendix 2) identifies one contributor (Organization Vivendi) to cause this outlier. This organization contributes to 2 candidates (Boxer, and Leahy) and the amount of abnormal accruals is - $ 8429,028.
Since all of the other abnormal accruals remain in a range from larger than - $ 5000, and smaller than $ 5000, this filter will be applied for the remaining research to prohibit outliers to influence the overall model. This results in a total of 214 contributions to be included in the sample.
Table 3 provides a summary of statistics regarding the sample on which the estimated regression will be ran. Data is retrieved for the year 2004 for the election candidates on
23
Table 3
Descriptive Statistics for Variables Influencing Election Outcomes
Variables N Minimum Maximum Mean Std. Deviation
Abnormal Accruals 214 - $ 4,582 $ 3,705 $ 39.01 $ 1,270.171
Standardized Abnormal Accruals 214 - 3.638 2.886 0.000 1.000
Incumbent 214 0 1 0.730 0.446
Republican 214 0 1 0.610 0.489
Valid N 214
whether they are Incumbent or not, whether they are Republican or not, and on the abnormal accruals per contribution.
As Table 3 shows, the mean of the abnormal accruals from the contributing firms is $ 39.01. Given that the abnormal accruals are calculated as the total accruals in 2004 minus the total accrual in 2003, an increase in the use of total accruals can be concluded for the election year 2004. Thus, supporting the finding of Ramanna & Roychowdhury (2010) that firms use more accruals in years of elections, since this is a positive figure.
Furthermore, the minimum and maximum figures for the variable Standardized Abnormal Accruals indicate that the correction for outliers has been applied correctly. This can be stated, because the Z-Score of the abnormal accruals provides an indicator of outliers. An outlier can be identified using the standardized values when a standardized value has a Z-Score above 4 or below - 4. As Table 3 shows, this is not the case.
4.3 Analysis & Results
In this section, the regression will be run over the final sample of 214 contributions and 50 candidates. The dependent variable is the dummy win, and the probability of the distribution is binominal. For this model, the link function will be logit. First, the results from running the empirical model are presented in Table 4 (correlation Table) and Table 5 (regression Table). Secondly, these results are analysed and explained. The main findings and arguments are stated in this subsection, and they form the basis for the final conclusion in paragraph 5.
4.3.1 Correlation Analysis
Firstly, the correlation between the variables that are expected to cause the main effects on the probability of a candidate to win or lose the election are presented and discussed. Table 4 shows the correlation between each of the variables with one another.
24
Table 4
Correlation Table for Main Effects (N = 214)
Variables 1. Abnormal Accruals 2. Incumbent 3. Republican
1. Abnormal Accruals - 2. Incumbent 0.010 (0.890) - 3. Republican 0.051 (0,457) -0.189* (0.006) -
* Correlation is significant at the 0.01 level (two-tailed).
The correlation coefficients of Table 4 indicate that Incumbent and Republican are negatively and significantly correlated. This implies that relatively more Democratic Senate members are up for re-election, as compared to Republican Senate members. This negative correlation was unexpected. But an explanation for this correlation can be found that the elections of the Senate are hold every six years, and by the elections of 1998 relatively more Democratical candidates won the Senate races (McLaughlin, 2014). This results in more Senate members being a Democrat and therefore being able to put themselves up for re-election.
Also, the correlation between Incumbent and Abnormal Accruals is positive, but not significant. In addition, the correlation between Republican and Abnormal Accruals is also positive, but not significant.
In overall can be stated that the correlations of the variables are quite low, since their absolute values all remain above the 0.00 and below 0.300. This range means that there barely is any correlation between the variables (Tilburg University, n.d.), which is positive for the empirical model which has previously been estimated. This is the case, because a high correlation between the variables negatively affects the reliance and relevance of outcomes of the regression model. The underlying reason beyond this, is that the outcomes of the regression could be caused by the interrelatedness of the variables. Therefore, the ability of the model to generate information that is applicable to a broader setting than the research sample declines, which in turn negatively affects the informativeness of the findings. Having established this, the regression can now be ran on the sample.
4.3.2 Regression Analysis
For the empirical model, an hierarchical regression is provided in Table 4. The hierarchical regression involves the interaction between categorical scores (being the dummies Incumbent, and Republican, and a continuous score (being Z Abnormal Accruals). The regression table
25
Table 5
Hierarchical Regression Analysis for Variables Influencing Election Outcomes (N = 214)
Variables Predicted
Sign
Model 1 Model 2 Model 3
Z Abnormal Accruals + -0.015 (0.027) 0,582 0.246 (0.261) 0.345 0.276 (0.261) 0.291 Incumbent + -3.378** (0.459) 0.000 -3.523** (0.493) 0.000 Republican +/- 0.508 (0.490) 0.300 Log Likelihood -107.789 -59.133 -62.300 R-Square 0.001 0.377 0.380
Likelihood Ratio Chi-Square 0.303 74.620 75.719
This table reports the regression results of the following equation:
𝑦 = 𝛽0 + 𝛽1 𝑍 𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝐴𝑐𝑐𝑟𝑢𝑎𝑙𝑠 + 𝐶1 𝐼𝑛𝑐𝑢𝑚𝑏𝑒𝑛𝑡 +/−𝐶2 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 + 𝜖
where the first value represents the value of β, respectively C, the second value represents the standard error of the β, respectively C, and the last value represents the level of significance. The levels of significance are based on the variables being two-tailed, and are indicated in the Table with *, and **, which represent p < 0.05. and p < 0.01 respectively. Log likelihood and likelihood ratio chi-square are measurements for the overall model. Log likelihood is estimated by Goodness of Fit test, and the likelihood ratio chi-square is estimated by an Omnibus Test.
shows the results for 3 different models, from which Model 3 captures the complete empirical model. The regression table enables me to show how adding each variable to the regression resulted in an overall stronger regression model. For example, adding the control variable Incumbent improved the overall model, since it resulted in the Log Likelihood becoming closer to zero, and it increased both the Likelihood of the Ratio Chi-Square and the R-Square ratio.
Although adding the other variables one by one into the model shows that the overall model, and the contribution of the variables to the model become more significant, the most significant effect was expected to be caused by the standardized abnormal accruals in accordance with hypothesis 1. However, this appears not to be the case. The significance level of the abnormal accruals started in Model 1 at a not significant level of 0.582. It is not after adding the variable Incumbent to the regression that the significance level of the abnormal accruals becomes closer to zero. However, even in Model 3 the standardized abnormal
26 accruals remain largely insignificant. Moreover, the standard error of the standardized abnormal accruals have increased while adding more variables to the model, as well as the steepness of its β.
Thus, it can be stated that the effect of the standardized abnormal accruals alone on the election outcomes are not significant. It appears that the variable Incumbent significantly influences the significance of the effects of the standardized abnormal accruals on the election outcomes. Also, the variable Incumbent seems to smoothen the effect of the abnormal accruals on the dependent variable Election Outcome, by increasing the value of the β, and the standard error of the β. From this can be concluded that the abnormal accruals of contributing companies in years of elections do not influence the probability of a candidate to win or lose an election. This leads to the rejection of hypothesis 1, so electable candidates do not have a greater (weaker) chance to win an election if a corporate sponsor uses more income decreasing (increasing) accruals in years of elections.
Also, in this setting it also seems that the variable Incumbent is highly significant itself, since it’s p-value is smaller than 0.01. So, opposed to the variable Standardized Abnormal Accruals, the variable Incumbent does affect the outcomes of elections significantly. Although this is consistent with the prior literature, the way in which it affects the chances of a candidate of winning differs significantly from the prior literature. The expected effect of a candidate being Incumbent on the election outcome is positive. However, the results in this setting show that being Incumbent has a negative effect on the changes of a candidate to win a race throughout Models 2 till 3. The steepness of the variable in Model 3 is -3.523, which leaves room for further research to inquire the reasons behind this negative effect. It could be caused by some specific factors surrounding the elections of 2004, which were not included in this empirical model.
Lastly, adding the control variable Republican to the model results in a decrease of the Log Likelihood from -59.133 to -62.300. However, the Likelihood Ratio Chi-Square increased from 74.620 to 75.719, and the R-Square also increased slightly from 0.377 to 0.380. These contradictory effects on the quality of the overall model are difficult to interpret. But since the significance of the standardized Abnormal Accruals increased with 5 percent (from 0.345 to 0.291), without altering the standard error of the estimate, and both the R-Square and the Likelihood Ratio Chi-R-Square increase slightly, I presume that the overall model has increased in quality. This implies that the overall model has increased in quality, and thus in its ability to explain the dependent variable.
27 To summarize the analysis, the findings of this analysis indicate that the abnormal accruals of organisations that contribute large amounts of money to Senate candidates does not affect the chances of these candidates to win or lose the elections. Also, in this setting it appears to be the case that being a candidate who is up for re-election has a significant negative effect on the probability of winning the election. Moreover, the results from the regression model indicate that the variable incumbent seems to weaken the effects of the abnormal accruals on the election outcome.
28
5. Conclusion & Discussion
19.26 Percent of the organisations in America spend money on direct and indirect lobbying. This has been calculated by Correia (2014) based on the current known ways of lobbying. However, there might be an additional way to gain influence in the decision making process of America. Ramanna & Roychowdhury (2010) found that organizations with a history of sponsoring political candidates tend to use more income decreasing accruals in years of elections to avoid political costs and scrutiny over their favoured candidates. They further state that this could be a possibility for those organisations to gain influence in the American politics, since the avoidance of scrutiny over the sponsored candidates might lead to an increased chance of winning the elections by those candidates (Ramanna & Roychowdhury, 2010). This possibility has not been addressed in the current literature yet, as far as I am aware.
But, if this possibility indeed exists, it would seem that the political cost theory as defined by Watts & Zimmerman (1978) does not cover the complete motives of management to make political expenditures (lobbying). Also, if accrual adjustment provides a possibility for organisations to gain influence in the American Politics, society needs to be aware of this possibility and that their interests are no longer equally represented in the politics as is assumed in a democracy (Fisher & Eisenstaedt, 2004).
In order to investigate this matter, the following research question has been formulated: “What influence did the accrual choices of corporate sponsors have on the probability of Congress candidates to win the Congress election of 2004 in America?”. An initial sample of 33 Senate races with 66 candidates, 1327 contributions and a total contributed amount of $ 35,603,229 has been estimated to answer the research question. After combining data from Compustat on abnormal accruals for contributors with hand collected data from Opensecrets.org, and CNN.com, a final sample of 33 races, 50 candidates, and 214 contributions with a total amount of $ 5,413,004 remained.
Also, an empirical model has been developed to explain the dependent variable: election outcome. The variables included in the final model are: standardized Abnormal Accruals, Incumbent (being a control variable), and Republican (being a control variable). After running the empirical model on the final sample, the results indicated that there could be two outcomes of primary interest obtained from this research.
29 The first finding indicates that the abnormal accruals of contributors of the Senate candidates does not affect the changes of those candidates to win or lose the elections itself. So, based on this research it can be concluded that hypothesis 1 should be rejected. Thus, electable candidates do not have a greater (weaker) chance to win an election if a corporate sponsor uses more income decreasing (increasing) accruals in election years. Therefore, the answer on the research question is that it is unlikely that the accrual choices of corporate sponsors have an influence on the probability of Congress candidates to win the Congress elections of 2004 in America.
Furthermore, it seems unlikely that the political cost theory as defined by Watts & Zimmerman (1978) needs to be extended. For society at large, it is good to have estimated that the usage of accrual accounting techniques does not directly pertain organizations to gather influence in the American politics, since the interest of society are not endangered by the interest of organizations being overrepresented in the decision making process.
The second outcome is related to the control variable Incumbent, and implies that if a Senate candidate is up for re-election, this negatively affects his or her chance to win the race, as compared to a candidate who is a new challenger. This effect is opposed to findings of prior literature, however no explanation could be provided for the negative effect based on this research. Furthermore, the results from the regression model indicate that the control variable Incumbent seems to weaken the effects of the abnormal accruals on the possibility of a candidate to win a Senate election race.
A limitation of this study, is that the effects of both the positive and negative abnormal accrual figures of contributors are taken into account together. Perhaps there is a difference between the effect of increased and decreased usage of accruals in years of elections, which have not been investigated in this research. This limitation provides a possibility for further research to look deeper in the abnormal accruals of companies in years of elections.
Also, further research could try to search for the reason behind the negative effect of an incumbent candidate on his or her winning possibilities as compared to a new challenger. This finding was opposite to the expected effect of being incumbent on the election outcome, and my research has not been able to find a suitable explanation for this phenomenon.
Lastly, it could be interesting to look at the reason behind the changing effect of the control variable Republican on the outcome of the election. The regression table showed that adding the interaction variable to the model resulted in the steepness from the control variable Republican first being positive, resulted in being negative. I have been unable to explain this
30 change of value, but I expect that there are other factors surrounding the election of 2004 which have not been included in this research are responsible for this turn of the C value.
31
Reference List
Abramowitz, A.I. & Stone, W.J. (2006). The Bush Effect: Polarization, Turnout, and
Activism in the 2004 Presidential Election. Presidential Studies Quarterly, 36 (2): 141 – 154.
Alexander, H.E. (1989). Comparative Political Finance in the 1980s. Cambridge, England: Cambridge University Press, Chapter 1: Money and politics: rethinking a conceptual framework.
Alford, J. & Hibbing, J. (1981). Increased incumbency advantage in the House. Journal of
Politics, 41: 1042 – 1061.
Campbell, A., Converse, P. Miller, W.E. & Stokes, D.E. (1966). Elections and the Political
Order. New York, United States of America: John Wiley and Sons.
Chi, W., Lisic, L.L. & Pevzner, M. (2011). Is Enhanced Audit Quality Associated With Greater Real Earnings Management? Accounting Horizons, 25(2): 315 - 335.
CNN (2005). U.S. Senate / Complete Results. Retrieved January 19, 2015, from http://edition.cnn.com/ELECTION/2004/pages/results/Senate/full.list/
Correia, M. M. (2014). Political Connections and SEC enforcement. Journal of Accounting
and Economics, 57: 241 – 262.
DeAngelo, L. E. (1986). Accounting Numbers as Market Valuation Substites: A Study of Management Buyouts of Public Stokholders. The Accounting Review, 61: 400 – 420.
Fields, T., Lys, T. & Vincent, L. (2001). Empirical Research on Accounting Choice. Journal
of Accounting and Economics, 31: 255 – 307.
Fiorina, M.P. (1977). Congress: Keystone of the Washington establishment. New Haven, United States of America: Yale University Press.
Fisher, J. & Eisenstaedt, T. (2004). Comparative Party Finance: What is to be done? Party
Politics. Retrieved January 26, 2015, from http://www.researchgate.net/publication/
246524119_Introduction_Comparative_Party_Finance_What_is_to_be_Done
Green, D.P. & Krasno, J.S. (1988). Salvation for the Spendthrift Incumbent: Re-estimating the Effects of Campaign Spending in House Elections. American Journal of Political
