Mo’ money, mo’ votes: Does special interest money affect senators’ voting proclivities on climate change?

Noah Yosif

Abstract: Extravagant political spending (“big money in politics”) is an important issue in contemporary American politics. Many academics, policymakers, political scholars, and citizens believe that excessive campaign spending is counterproductive to fostering an equitable system of governance where everyone has equal access and representation by their elected officials. Central to this argument is the notion that exorbitant political donations persuade legislators to vote in a manner conducive to the interests of their donors. However, this assertion fails to account for many other variables – such as the party line, the officials’ political standing, and the issue itself – that would also influence a legislator. This paper employed a multiple regression model to assess the impact of special interest donations on United States Senators’ voting on climate change-related legislation. Results indicate that political donations do indeed influence a senator’s overall voting calculus on climate change policies. Although the effect on a per-dollar basis appears to be modest, it can be significant when scaled by the massive amounts donated. The quantitative methods developed in this study could also help inform our understanding of the influence on big money on other issues.

Keywords: United States Senate, climate change, electoral politics, money in politics

I. Introduction

Exorbitant campaign donations by wealthy individuals and corporations, aptly termed as “big money in politics,” is a dominant issue in contemporary American politics. Since the late 20th century, special interest donations to political campaigns have been increasing, yet this issue has attracted more scrutiny in recent years given the enormous size of some contributions and their efficacy in pushing the particular policy platforms of their donors (OpenSecrets.org, 2019). This can be attributed to a lax regulatory framework that has failed to keep pace with an American constituency characterized by income inequality among individual citizens as well as an economic environment where large corporations have embraced lobbying to further their business interests. In light of direct legal efforts to constrain limitless political donations such as Citizens United, as well as poignant issues such as gun control and abortion, which have highlighted special interests’ power over the policymaking process, “big money in politics” has indubitably become a divisive issue in (Avis et al., 2017). Setting aside the legal and moral considerations that arise from such an issue, this paper will attempt to answer an underappreciated economic question central to big money in politics: How effective are political donations in swaying a senator’s voting proclivities?

The answer is perhaps fairly obvious. Theories spanning human psychology (Eisler, 2016), political science (Primo and Milyo, 2006), and economics (Welch, 1974) agree that the odds of currying favor from an individual are much better when asked alongside a modest check. This is perhaps why comprehensive quantitative analyses of big money in politics are few and far between. The vast majority of available literature on big money in politics is often restricted to documenting the contributions of special interests on one particular issue and employing qualitative analysis to connect monetary contribution to passage or failure of legislation (Bowler and Donovan, 2016). This approach fails to quantify the impact of special interest donations on the inclinations of legislators, and the reductive assertion of more money equating to more influence negates other important factors, such as the issue itself, the senators’ personal philosophies, their party’s views, their political standing, socioeconomic consequences, and so much more. The inclusion of these variables could seriously complicate a supposedly simple relationship between donors’ money and their political influence. A more sophisticated quantitative estimation of a donor’s acquired influence over a legislator would be important to the public discourse on this issue, and more importantly, for potential regulatory efforts to reign in exorbitant campaign donations.

This paper will estimate the impact of special interest donations on legislators’ voting proclivities by focusing on one issue, climate change, and developing a multiple regression model that examines the impact of increased contributions on convincing U.S. Senators to promote a particular policy agenda. Climate change is a good issue with which to analyze the impact of big money in politics because it is divisive and attracts significant attention from wealthy special interests on both sides of the political spectrum. The model will synthesize several political, economic, and climate change-specific variables that would likely be part of the calculus of a legislator’s stance on climate change-related bills. The model will attempt to quantify the impact of political donations on senators’ overall voting rationale. This paper hopes the model and resulting analysis could contribute to the ongoing conversation on climate change, and more broadly, the impact of big money in politics. 

II. Literature review

            As mentioned above, the available literature covering big money in politics is diverse but fails to adequately assess the impact of campaign contributions on a legislator’s likelihood of aligning their votes with their donors’ interests. The notion of campaign contributions equating to political influence is popular: A study conducted by Justin Grimmer and Eleanor Powell (2016) found that an overwhelming majority of Americans who disapprove of Congress named “corruption” of the political process facilitated by excessive, unregulated campaign contributions as their main grievance. While average citizens believe big money in politics is a problem, donors see it as a solution and viable vehicle for realizing their political views. Alexander Fouirnaies and Andrew Hall (2015) found that firms within highly regulated industries are more likely to donate money to incumbent legislators, believing greater contributions preserve the status quo and facilitate continued access. If a legislator is particularly powerful because they hold a committee position or occupy a major leadership role in their party, donors are likely to contribute more money to them than they would to a challenger or freshman legislator.

Yet, the majority of available literature also acknowledges the difficulty of establishing a causal link between campaign contributions and legislators’ voting proclivities. Janet Grenzke (1989) argues that endogeneity is the main problem when analyzing this issue quantitatively. She maintains that a correlation between increased donations and a senator’s voting record may be spurious (Grenzke, 1989). For example, if a pro-climate legislator met with a pro-environment interest group which also donated to their campaign, a link between donations and political access may appear to exist even though the actual explanation for the meeting stemmed from mutual political goals. Nathaniel Persily and Kelli Lammie (2004) argue that omitted variable bias is the primary obstacle to establishing a causal relationship between political contributions and legislators’ behavior. For example, if there is a strong public stigma attached to a particular political position, such as funding recovery measures after a major oil spill, campaign contributions may not be enough to sway a legislator’s vote (Persily and Lammie, 2004). These studies demonstrate how any empirical assessment of political contributions on politicians’ behavior would yield uneven results that are difficult to generalize.

In spite of these challenges, several studies have provided glimpses into the power of political donations on legislators’ votes and the legislative process. Joshua Kalla and David Broockman (2016) analyzed the impact of campaign contributions on preferential treatment from policymakers through a randomized field experiment. In this study, CREDO Action, a progressive political organization, made 191 political donations, randomly assigning whether it revealed to congressional offices its status as a contributor. It found that senior-ranking[1] legislators were three to four times more likely to be available to meet and take steps to address their concerns (Kalla and Broockman, 2016). Another study by Justin Grimmer and Eleanor Powell (2016) found that corporations and business political action committees (PACs) donated significant amounts of money to incoming congressmen on committees that had regulatory power over their particular industry but decreased or ended contributions to committee members who were retiring or stepping away from their post. This study showed that business interests seek short-term access to influential legislators, and they believe their donations will have a favorable impact on the policy process.

Additionally, several studies have analyzed the influence of political donations on climate change policy. A study by Justin Farrell (2016) at Yale University concluded that legislators who received the most political donations from special interests were likely to also have the most polarized views on the matter in favor of their donors. While this study mentions political contributions to come from donors on both sides of the climate change issue, it does not give any details to the profile of a pro-climate or anti-climate change donor. Robert Brulle (2018) answers this question by analyzing climate change lobbying trends from 2000 to 2016. He concludes that the majority of corporate or PAC-based political contributions come from donors opposed to additional climate change measures, while individual contributions are almost evenly split between donors on both sides of the issue (Brulle, 2018). Ans Kolk and Jonatan Pinkse (2017) delve further into this question by examining the demands of corporate donors with regards to climate change and find their political activities can be characterized as an information strategy to steer policymakers toward market-based solutions as opposed to increased regulation.

The available literature showed that individuals, corporations, and policymakers believe in the power of political donations to effect policy change on climate change. However, the exact influence of big money is hard to quantify given the confluence of other variables affecting a legislator’s calculus, such as their thoughts on the issue, public sentiment, and competing interests. The literature also indicates that donors get more attention from legislators and have a greater likelihood of shaping the political discourse around a particular policy by pushing more money toward individuals with the highest authority or regulatory power. Finally, the existing literature describes how campaign contributions have polarized the debate around climate change and provided a glimpse into the profiles of pro-climate and anti-climate change donors. There has been much academic progress studying both big money in politics and climate change, and this study will bridge the two issues together by quantifying the impact of political contributions on legislators’ voting proclivities on climate change.
 

III. Methodology and data

A multiple linear regression model is the preferable method for quantifying the impact of campaign contributions on legislators’ voting proclivities. For this study, an ordinary least square (OLS) method of estimation was employed for time-series data between 1990 and 2016. A legislator’s opinion expressed through voting can vary significantly due to a myriad of considerations. Limiting the study to one issue such as climate change makes the analysis monumentally easier because it reduces the quantity of potential variables to include in the model. To further narrow the scope of this study, the term “legislator” is defined as a United States Senator. Similarly, the term “voting record” will describe a ratio of pro-climate change votes to total votes cast on climate change legislation. The raw data utilized in constructing this variable rated every senator based on whether they voted in favor of climate change-related policy, against climate change-related policy, or did not vote at all (absence or abstention). Senators who were absent and did not vote were excluded from the construction of this variable to correct for outliers and avoid skewness in data.

The first group of variables included in this study encompasses various political considerations that would be typical for any legislator. These are a senator’s voting record on climate change issues, a senator’s political party, the party in control of the Senate, and the party in control of the executive branch. While the dynamics of these variables are specific to the climate change issue, they are also critical questions that would be considered by most senators when voting on legislation regarding any issue. The next group of variables are economic indicators, which include GDP, inflation, the unemployment rate, and instance of recession. These are particularly applicable to the subject of climate change since much of the policy debates surrounding this issue are rooted in a country’s economic health. For example, large swaths of voters in France opposed recent energy tax hikes to make the country more environmentally friendly because of inflation alongside skyrocketing costs of living (Goldhammer, 2018). The final group of variables is specific to the issue of climate change, including land surface air temperature, an internationally recognized indicator of climate change, and public interest in the matter. Each of the variables used for this study is listed below: 

Senator’s voting record: This is a ratio of every senator’s pro-climate change votes to all votes cast on climate change-related legislation per year, compiled using 2017 data from the League of Conservation Voters (CLV). This is the dependent variable.

Senator’s political party: This is a dummy variable compiled from 2017 LCV data on each senator’s voting record, including years individuals switched parties or retired. The variable is 0 for Democrat and 1 for Republican.

Control of the Senate: This is a dummy variable that shows which party holds the majority of seats in the United States Senate. The variable is 0 for Democrat and 1 for Republican.

Control of the executive branch: This is a dummy variable that represents the political party of the President of the United States. The variable is 0 for Democrat and 1 for Republican.

GDP: This is a continuous variable that measures the percent change in real GDP. This data was accessed through the St. Louis Federal Reserve Bank Economic Data (FRED) database. 

Inflation: This is a continuous variable and represents the annual percent change in the cost of a basket of goods and services; it was accessed through FRED.

Unemployment rate: This is a continuous variable and a measure of the number of unemployed as a percentage of the overall labor force. This data was accessed through the FRED.

Instance of recession: This is a dummy variable that represents whether there was a significant decline in economic activity that lasted longer than several months. The variable is 0 for no recession and 1 for the presence of a recession. The data was accessed through the National Bureau of Economic Research.

Contributions from climate change interests: This represents a dollar figure of climate-related special interest donations to senators above $200, as compiled by the Center for Responsive Politics. According to the Federal Election Commission, campaign contributions aggregating to over $200 are considered large donations and subject to specific reporting requirements (OpenSecrets.org, 2018). This is the primary independent variable of the model.

Land surface air temperature: This is a measure of actual climate change and the only variable that is an indicator of climate change relative to senators’ votes on policies that address it. The data was retrieved via the National Oceanic and Atmospheric Administration (2016).

Public interest: This is a percent change in public attention toward climate change as an issue. It was collected from a study by researchers at Princeton and Oxford University that employed a publicly available dataset of worldwide web search term volumes to detect temporal patterns of interest in climate change between 2000 and 2014. For years outside this time period, a three-year moving average method of estimation was applied (Anderegg and Goldsmith, 2014).

In all, this model is comprised of four discrete variables, four dummy variables, and three continuous variables. There were 10 observations where senators’ political contributions or voting record were unavailable; these estimates were removed from the dataset to prevent the sample statistics from skewing. Figure 1 is a graphical display of the raw data for six of the 11 variables within the model; senator’s voting record was excluded because it is a ratio that varied per senator, and data aggregation techniques would have obscured the values too much. There was a 106% spike in climate change-related campaign contributions in 2008, but this could possibly be explained by the timing, since 2008 was a contentious election year for both the presidency of the United States and the Senate. GDP and the unemployment rate smoothly fluctuated from year-to-year, and the latter appears to move in tandem with inflation.

Spikes in the unemployment rate from 1990 to 1992, 2000 to 2002, and 2008 to 2010 align closely with the directional movement of inflation. Land surface air temperature doubled over the time period, while public interest in the subject remained constant, but never dipping below 50%. While the data behind this variable was taken from an academic survey, it shows that a median of around 60% of people (and at times as high as 78%), were at the very least attentive to the climate change issue, showing it to be an important topic in American political discourse. By applying an Augmented Dickey-Fuller (ADF) test, we confirmed that all of these variables are stationary. 

Figure 2 shows the results of the ADF test for all 11 variables. For this test, the null hypothesis is the presence of a unit root that the observations within a dataset coalesce around, otherwise known as non-stationarity. The alternative hypothesis is stationarity, which is the desirable outcome indicating that the statistical properties (variances and means) of the dataset do not change with time. Each of these variables posted a p-value of 0.01, thereby indicating stationarity. After determining the stationarity of each variable, the next step is to test for cointegration. For this test the null hypothesis is no cointegration within the time series while the alternative hypothesis is that cointegration is present. The results are displayed in Figure 3. Since the p-value is below 0.01, the null hypothesis is rejected, indicating that presence of cointegration cannot be rejected. These newfound properties of stationarity and cointegration do not completely eliminate the possibility of a spurious relationship between the dependent and explanatory variables, which means further analysis must be conducted.

III. Results and analysis

Many different models were tested to determine an accurate relationship between senators’ voting records and the donations they received from climate change special interests. These models included different combinations of the 11 independent variables mentioned above, including lagged or transformed versions of them. Cointegration was a significant issue and was particularly prominent within the economics-related variables: GDP, inflation, unemployment rate, and instance of recession. This is to be expected, as these variables have well-known relationships. During instances of recession, GDP decreases and when there is no recession, it increases. As per the Phillips Curve, the unemployment rate decreases as inflation increases and vice versa. Control of the senate and control of the executive branch had the highest fluctuating significance; when combined with some variables, they had a high explanatory power and when combined with others they were relatively insignificant. While these two variables do not have any obvious correlation with any other variables within the model, there is reason for them to be cointegrated. If one party controls the Senate and executive branch, their legislative power is significantly enhanced which would make donations to senators more insignificant than if power was split between two parties controlling one policymaking apparatus each. In light of this, subsequent models contained either control of the senate or control of the executive branch as variables, but not both. The rest of the analysis will focus on the results of Figures 4 and 5, which highlights the best model.

A mathematical representation of the final model is also shown below:

where Y is the dependent variable for Senator i in time t;  is a vector of political controls that includes senator’s political party and control of the Senate;  is a vector of economic conditions, including the unemployment rate and instance of recession; T refers to the global land surface air temperature; I represents public interest; and  is the error term.  is the key independent variable and represents donations by interest groups.

In this model, contributions from climate change interests, senator’s political party, control of the Senate, the unemployment rate, instance of recession, land surface air temperature, and public interest are all significant variables that determined a senator’s voting record on climate change issues from 1990 to 2016. This model sports a 0.74 R-Squared employing a total 1,350 observations. It is essential to note that a senator’s voting record is the percent of climate change-friendly votes on pending pieces of legislation; an increased/generally high voting record indicates the senator will vote for climate change policies and vice versa. All else constant, the following observations can be made from this model:

• A senator’s political party significantly affects their vote on climate change issues; specifically, since this is a dummy variable (0 for Democrat and 1 for Republican), the regression indicates that Republicans are 65% less likely to vote for climate change-related matters than Democrats.

• A one-unit increase in the land surface air temperature is associated with a 16% increase in a senator’s voting record. This is as expected because higher temperatures associated with climate change would be disconcerting to voters who would be expected to persuade their senator to vote in favor of climate change policies.

• A 1-unit increase in public interest is associated with a 93% increase in a senator’s voting record, as this is the most direct variable for connecting public interest to a senator’s proclivities on an issue. The more a senator’s constituents advocate a particular issue, the stronger the likelihood they will vote in favor of it.

The most important observation within this regression is that each dollar in contributions from climate change interests is associated with a decrease of 0.00000001% in a senator’s proclivity to vote in favor of climate change policies. While modest on a per-dollar basis, the relationship could be significant for large donations from multiple donors in a hotly contested election year. According to OpenSecrets.org, the average amount of donations given to senators from climate change-oriented special interests in 2016 was $185,000, which is associated with an average of 2% decrease in a senator’s voting record. The effect of a modest 2% change in a senator’s voting record is put in better context when considering the number of climate change-related votes in the Senate. In 2016, there were 17 pieces of climate change-related legislation that received a full vote in the Senate; the five-year average was 14. Therefore, a 2% difference could potentially have a major impact by costing votes on individual pieces of legislation.

Figure 6 examines the 2016 average, as well as the first, second and third quartiles of data to illustrate how this model would estimate a senator’s voting record on climate change-related legislation in 2016. On average, senators are likely to vote less on climate change-related legislation if they are on the higher end of the campaign contributions spectrum.

As mentioned earlier, there were concerns of a spurious relationship between a senator’s voting record and the donations they received from special interest groups. This would mean that the dependent variable, a senator’s voting record on climate change, would be causally related to the donations they receive from climate change interests but is only by coincidence or due to a lurking variable. The chances of this scenario were already unlikely because the variables were proven to be stationary through via an Augmented Dickey-Fuller Test; however, cointegration was detected via the Phillips-Ouliaris Test. Figure 7 shows the ACF of the residuals, showing them to have near-perfect white noise and to not be correlated. Therefore, the results from this regression show that political donations to a senator have an impact on their votes toward climate change-related legislation, but this is not a casual or by-chance relationship.

IV. Limitations

One limitation of this paper was the institutional bias of the dependent variable – senators’ voting records on climate change-related legislation. This paper used data on senator’s votes on climate change from the League of Conservation Voters, which advocates for policies to combat climate change. Alternative measures of senators’ voting records on climate change-related issues were not readily available.

There are also a few caveats worth noting about the public interest variable. First, public interest – the degree to which he public is interested in an issue – is different from public opinion – what the public actually thinks about an issue. Furthermore, there is disagreement on the effectiveness of discerning public interest using Google searches issues. The merits of employing online searches as an indicator for Public Opinion have been hotly contested in academic literature, as well as whether such opinions are driven by elite partisan messaging from the media or influential public figures (Gramlich, 2017).

Finally, next steps for further analysis might entail replicating these experiments for the United States House of Representatives, as well as state legislatures. While including multiple legislative bodies in this experiment would have complicated the results of this study since they are affected by different politics, building separate models which equitably test their susceptibility to political donations may enhance the external validity of this study.

V. Conclusion

Big money in politics is a poignant issue in contemporary American political discourse. According to a Pew research poll, 77% of Americans believe that big money has an impact on the legislative process, and stronger laws would ensure a more equitable system (Jones, 2018). In spite of significant public attention to this issue, there has been little academic study to assess quantitatively the effect of big money on driving political change via legislators’ voting proclivities. This paper attempted to supplement this important discussion by offering a glimpse into the effect of big money in politics through the lens of a single issue, climate change. The paper determined that political donations have a significant influence on a legislator’s voting proclivities on climate change. Legislators are more likely to vote in the interests of their donors, which affirms that deep pockets are an effective vehicle for driving legislative change. Having ascertained political donations to have a major impact on legislators’ voting proclivities within the realm of climate change, one can only wonder if our elected officials on Capitol Hill would change their minds on gun control, healthcare, and abortion for only a few dollars more.