The wrath of voters – what they do and do not
prioritise at the ballot box
A closer look at the impact of skipping votes in the U.S. Senate
Bachelor’s thesis Political Science
Name: Dominic Mostert
Student number: 1853937
Bachelor: Political Science
First reader: Dr. T.A. Mickler
Second reader: Dr. S.P. Otjes
1
Introduction
The ability to vote remains a fundamental feature of democratic systems. After citizens vote representatives into office, those representatives are expected in turn to vote on legislative matters, as is required in a system of majority rule (Dahl, 1989). All representative democracies feature this 'chain of delegation' from voters to elected officials, as well as a 'chain of accountability' that runs in the opposite direction (Strøm, 1999). The relationship between constituents and their representatives is one of mutual dependency, and hence, has been understood as a two-way street. Constituents yield significant power to their representatives and will usually have to wait until the next election to hold them accountable. For representatives, the next election is important, too: once in office, they mainly seek to be re-elected, as all their subsequent political goals are contingent on staying in office (Mayhew 1974; Fiorina 1989).
To better understand the dynamics of the relationship between voters and their representatives, it is crucial to understand how voters vote. Past research suggests that voters are capable of voting retrospectively (Key Jr, 1966; Persson and Tabellini, 1997; Canes-Wrone et al, 2002). “Voters are not fools,” Key Jr. stated (Key Jr., 1966, p. 7); they consider past performance by incumbents when making their decision at the ballot box. They make an active and personal judgment of an incumbent’s behaviour while in office. To do so, voters need information about the way their representatives have conducted themselves. That information can derive from various sources, but one of the most obvious sources is an incumbent’s voting record. By evaluating how their representative voted on the issues at hand, a constituent can determine whether their personal beliefs line up with those of his or her representative.
However, tension can arise when legislators display different levels of activity in their representative bodies. Requirements regarding activities that legislators undertake are usually not enshrined in a legislative framework; representatives enjoy leeway in their time allocation. They are not obligated, for example, to attend or engage in parliamentary debates, to submit parliamentary questions or to vote on new laws. Inevitably, therefore, some representatives engage more than others in these activities. In this paper, attention will be paid specifically to a lack of participation by legislators in roll-call voting. Rosas and Shomer (2008) tracked absenteeism in several different democratic legislatures throughout the world. In the seventeen most recent sessions of legislatures that they studied, ranging from 1984 until 2004, the average rate of missed votes constitutes 23.7 percent. Legislators abstaining from voting for or against
2
a proposition, an option allowed by some but not all legislatures, are not included in this number, as explicitly voting “present” is seen as a positive act of voting (Rosas and Shomer, 2008). The authors selected six legislatures in North America (Canada, Guatemala and four sessions of the U.S. Congress), five legislatures in South America (Argentina, Brazil, Chile, Ecuador and Peru), three in Europe (Czech Republic, Poland and the European Parliament), two in Oceania (Australia and New Zealand) and one in Asia (Israel). To calculate the average mentioned above of 23.7 percent, five sessions of the U.S. Congress were removed, as the U.S. would otherwise have an undue influence on the average and because those sessions were more dated (ranging from 1915 to 1978) than the other legislatures studied. While the selected legislatures did not constitute a random sample from the global population, the authors argued there was no reason to suspect that these assemblies exhibit abnormal rates of vote-missing. Voters depend on their representatives to vote on their behalf, but representatives have been said to have a ‘home style’ (Fenno, 1978), leading to differences in how constituents are represented by their representatives. Those who miss votes, however, run the risk of appearing irresponsible. Voters can choose to address this disparity in the next election, and there is reason to suspect that they punish legislators with a lax voting record. For example, Salmore and Salmore (1989) refer to current U.S. Senate Majority Leader Mitch McConnell’s upset victory over incumbent Senator Walter Huddlestone in 1984. McConnell argued that his opponent was lax in participating on roll-call votes, a line of attack that was considered effective at the time (Salmore and Salmore, 1989). A campaign ad featuring a hunter and blood hounds, searching throughout the country for Huddleston, was popular with viewers (Payne and Baukus, 1985). Other incumbents took note: Mintz (1996) states that the fallout from McConnell’s victory “seemingly precipitated roll call participation levels in both the House and the Senate to increase by about 3 percent.”
Incumbents are attacked frequently by challengers on their substandard roll-call participation (Brown and Goodliffe, 2017). However, the electoral consequences of absenteeism have not been studied much. Doing so can enhance our understanding of the relationship between voters and representatives in a representative democracy. This paper aims to broaden our understanding of the dynamic between voters and representatives, by asking: what is the effect
of roll-call absenteeism on electoral success of incumbents? In this paper, several hypotheses
are developed in a discussion of the existing literature on the determinants of re-election support. These hypotheses concern factors that voters are deemed to prioritise when deciding whether or not to vote for an incumbent. They will be tested through an analysis of the last
3
three Senate elections in the United States. After discussing the operationalisation of the different variables, a linear regression model is used to determine the effect and statistical significance of the factors involved. Voters do not appear to punish incumbents for missing a high number of votes. They appear to be guided more strongly by their party identification and their opinion of the executive branch.
Literature review
Voters are understood to have two ways of ensuring desirable policy outcomes. They can hold politicians accountable for their conduct in office and they can select politicians whose interests are closely aligned with those of the public. The former approach revolves around re-electing politicians that produced "good" outcomes and voting them out of office for producing “bad” ones. When citizens use their vote to retroactively hold politicians accountable, elections are viewed as mechanisms to ensure democratic accountability (Ferejohn 1986; Key 1966). The latter approach, however, results in elections functioning as mechanisms for democratic selection (Downs 1957; Fearon 1999; Manin 1997).
Previous studies on the determinants of re-election support the assumptions that voters vote retrospectively (Key, 1966; Persson et al, 1997; Canes-Wrone et al, 2001). Conducting a lab experiment testing the predictions of a game theoretic model, Woon (2012) finds that subjects’ voting behaviour is "consistent with a purely retrospective reward-punishment strategy" (Woon, 2012, p. 914), challenging theories of elections as selection mechanisms (Alesina and Rosenthal 1995; Ashworth 2005; Ashworth and Bueno de Mesquita 2008; Fearon 1999). Woon’s results align closely with empirical literature on economic voting (Kramer 1971; Nordhaus, 1975; Norpoth, 1996; Nadeau and Lewis-Beck, 2001).
Voters’ ability to punish incumbents is dependent on the electoral system of the country involved. In multi-party political systems, voters enjoy the possibility to punish or reward incumbents of several governing parties. In more restrictive political systems, specifically in two-party systems, voters will shift to the non-governing established party, even if that party was once deemed responsible for an economic decline (Benton, 2005). The state of the economy is one of the several factors distinguished in the literature affecting voting. Previous research indicates the existence of economic voting in largely stable economic environments like the United States and Western Europe, but also in Africa, Asia and Latin America (Pacek and Radcliff, 1995; Posner and Simon, 2002; Benton, 2005). Specifically in Latin America, Benton (2005) shows that voters have developed ‘long, sophisticated economic memories’, and
4
actively hope to punish incumbents of multiple parties for their roles in the economic crises of the years prior. Even incumbents from parties currently not in power are held accountable if they are deemed, at least partly, responsible for the economic downturns. In stable economies, on the other hand, voters need only consider short-term macroeconomic conditions (Benton, 2005). In Benton’s study of thirteen Latin American countries, incumbents lose 1.7 percent of the national votes for every percent decline in GDP per capita; for every percent of GDP increase, non-incumbents can expect to lose between 1.3 and 4.4 percent of the national votes. Also, voters have been shown to respond to increases and decreases in the unemployment rate. For example, studies in Turkey and the United States have shown that incumbent politicians are held accountable for heightened unemployment (Çarkoğlu, 1997; Burden and Wichowsky, 2014). These findings result in two hypotheses:
H1: An increase in GDP per capita during an incumbent’s term in office has a positive effect on the vote share for incumbents.
H2: An increase in unemployment during an incumbent’s term in office has a negative effect on the vote share for incumbents.
Besides the general state of the economy, party identification and ideological orientations influence voters’ decisions. Party identification refers to how strongly people identify with a specific political party, and denotes a psychological link towards that party (Campbell et al, 1960). Ideology is more fluid, refers to a set of beliefs, opinions and values, and is often measured through two-dimensional scales such as a liberal/conservative or left/right scale (Jost, 2017). Voters are looking for compatibility between themselves and their representatives and their goal is to have a government that reflects their ideological preferences (Pearse, 2005). Ideological compatibility between voters and incumbents is an important predictor of vote choice (Abramowitz, 1980, 1988; Abramowitz and Segal, 1992; Kahn and Kenney, 1997; Krasno, 1994; Westlye, 1991; Wright and Berkman, 1986).
Incumbents who deviate too much from their voters’ ideological preferences can expect tough re-election battles. Kau and Rubin (1979) and Kalt and Zupan (1984) argue that legislators frequently engage in ideological shirking, casting votes contrary to the interests of their constituents. Lott and Reed (1989) introduce the concept of 'finite-livedness', predicting that politicians are eventually 'sorted out' by their constituents, if their policy preferences differ from those of their constituents. In other words, ‘shirkers’ are likely to be punished by their constituents eventually. Those who shirk the least enjoy a greater chance of acquiring longevity
5
as a representative (Lott and Reed, 1989). Narrowing his focus on the United States, using data from 1964 to 1968, Wright (1993) finds that a higher level of ideological shirking by Senators indeed corresponds with a lower vote share in the subsequent election. These findings lead to two new hypotheses:
H3: A higher percentage of voters identifying with a party has a positive effect on the vote share of incumbents from that party.
H4: A higher level of ideological compatibility between incumbents and their voters has a positive effect on the vote share of the incumbent.
In presidential systems, voters’ assessments of the president are a relevant factor and touch on the presence or the absence of ‘presidential coattail effects’. Popular presidents can motivate voters to the ballot box, where they are more likely to vote for congressional candidates of the same party. If the president is not on the ballot, however, his or her party is more likely to suffer electoral defeats (Campbell, 1997). Especially with a controversial president in office, voters are eager to punish candidates from the president’s party in congressional elections (Campbell and Sumners, 1990). Research by Hinckley (1970) and Key (1958) also established the role of presidential coattail effects on congressional races in the United States. These results lead to a new hypothesis:
H5: In presidential systems, a negative assessment from voters of the executive branch has a negative effect on the vote share for incumbents from the same party.
Voters also remain susceptible to the outreach of well-financed campaigns. The influence of money on electoral prospects has been researched in several political contests, with scholars agreeing that money shapes the contours of electoral competition (Samuels, 2001). Well-financed campaigns have a better chance of succeeding. Samuels finds that in Brazil, if a candidate spends an additional percent of all campaign finance in a state, he or she may expect to add about half a percent to their total vote share. The results conform to research on campaign finance from the U.S, although institutional differences exist. Due to the unparalleled incumbency advantage in the U.S., deriving from a significant name recognition, additional spending suffers from diminishing returns. Financial prowess over one’s opponent still contributes greatly to electoral chances, but is not equally determinative in every country.
H6: Large financial support compared to a challenger has a positive effect on the vote share for an incumbent.
6
Voters have also been shown to respond to parliamentary productivity, although the results again vary by country. Bowler (2010) finds that voters reward members of parliament in the United Kingdom with a small but significant increase in total vote share for introducing private member bills. This effect is not observed in Israel by Akirav (2018), who points to Israel’s 'rather unusual and extreme' political institutions as a possible explanation, as well as a very weak connection between voters and their representatives in the country. The number of bills introduced is not the only way to measure parliamentary activity, as scholars realise that the outcome of the legislative process is perhaps even more important to voters. Cox and Terry (2008) choose to count the number of bills from each legislator that come out of committee, as well as the number of bills passed. While those variables will be included in the research, their electoral effects have not been studied as much. Therefore, only one new hypothesis based on the literature is added:
H7: The number of bills introduced has a positive effect on vote share for incumbents.
Additional forms of parliamentary activity exist that voters might show their appreciation for at the ballot box. The aforementioned factor of ideological compatibility implies that voters as well as politicians display ideological preferences. For incumbent politicians, their beliefs and values are often measured through their stances on call voting. Mayhew (1974) views roll-call votes as a signalling device used by legislators to communicate their stances on issues to voters. In this communicative process, only the signal matters – not the legislative outcome. Down’s (1957) analysis of the importance of ideology for parties supports the view that legislators will engage in roll-call voting when they deem sending a signal worthwhile in their electoral efforts (Cohen and Noll, 1991). Roll-call voting provides legislators with an opportunity to obtain a clear policy orientation, no matter the eventual result of the vote. Glazer and Grofman (1998) build on these findings, arguing that legislators who are perceived by their electorate to have clear ideological positions hold an electoral advantage, as voters find it easier to determine how they will act in the future. Prima facie, one would expect that missing a vote does not contribute to garner an image of having clear ideological positions, as those positions could be expressed better by either voting for or against a bill, or abstaining. A subpar attendance record has been theorised to have negatively affected incumbents’ re-election efforts. Voters might have punished U.S. Senator Walter Huddlestone in 1984, as mentioned before, and other possible examples will be mentioned below in the section on case selection. For now, the available literature leads to the main hypothesis:
7
H8: Missing votes has a negative effect on the share of votes for incumbents.
This paper aims to build on the existing literature, by attempting to determine whether participatory shirking yields the same results as ideological shirking. Lott’s and Reed's (1989) theoretical treatment of the issue predicts that voters will punish ideological shirkers, but it is unclear whether the same holds true for participatory shirkers.
Method and case selection
A case study, defined by Gerring (2004) as “an intensive study of a single unit with an aim to generalize across a larger set of units”, will be used to measure the effect of absenteeism on the electoral success of incumbents. Since a comprehensive study regarding the relationship has not yet been performed, the best way to measure the potential effect is to find a case that is most likely to show a relationship, if one exists at all. The most-likely case study adheres to the logic of the “inverse Sinatra inference" – “if I cannot make it here, I cannot make it anywhere” (Levy, 2002, p. 442).
The most likely case to display an effect would have to be a democratic system of government, as voters need an opportunity to hold their representatives accountable for their conduct in office. Also, a high level of personalised politics is desirable. Ideally, incumbents’ vote record, including their attendance rate, is part of the public discourse. If incumbents’ substandard attendance rates are publicly discussed, voters are more likely to be aware of them. For these reasons, the case selected is the United States Senate. The hypotheses will be tested through the lens of the 2012, 2014, 2016 and 2018 Senate elections.
There are several reasons to regard the U.S. as the most-likely case to display an effect. A democratic system of government is necessary to measure a potential effect, as voters need an opportunity to hold their representatives accountable for their conduct in office. In the U.S., members of both chambers of Congress are elected directly by voters. Elections are highly personalised; incumbents running for re-election are scrutinised by their opponents and their voting record is often successfully used against them in election campaigns (Capella and Taylor, 1992). Specifically, opponents are eager to point out high levels of absenteeism of the incumbents whose seat they are after (Brown and Goodliffe, 2017). When Representative Patrick Murphy challenged Marco Rubio for his Senate seat in 2016, Murphy was quick to point out that Rubio “has the worst vote attendance record of any Florida Senator in nearly 50 years” (Hagen, 2016). In 2018, when vying for Ted Cruz' Senate seat, Beto O'Rourke criticised Cruz' attendance record in the Senate. "You tell me, who can miss half the days at work and
8
then be rehired for the same job going forward? That is not what Texans want,” O’Rourke argued (Eversden, 2018). Similar attacks have been lodged at sitting members of the House of Representatives. In the run-up to the 2018 elections, The Tennessean levied criticism at Tennessee Representatives Marsha Blackburn and Diana Black for their substandard attendance record (Ebert, 2018). The South Dakota Democratic Party scrutinised Republican Representative Kristi Noem's voting record, accusing her of "neglecting" her responsibility to represent the state on the floor of the House (South Dakota Democratic Party, 2018). In 2012, Bill Nelson asked Representative Connie Mack during a debate to "explain how this year you have one of the worst voting records” (C-SPAN, 2012). Florida Congresswoman Debbie Wasserman-Schulz was, according to her opponent, “more concerned with headlining fundraisers for President Obama and the DNC than she is with fulfilling the responsibilities owed to her constituents” (Sherman, 2012). Representative Gwen Moore was dubbed “Wisconsin's most absent member of Congress, missing nearly 17 percent of the House votes in the second quarter of 2012” (Nelson, 2012).
Furthermore, the U.S. political system is dominated by two parties, with the Democratic Party and the Republican Party receiving a combined total of 97.2 percent, 96.2 percent, 95.5 percent and 95.8 percent respectively in the 2018, 2016, 2014 and 2012 Senate elections. Other parties generally play a marginal role in electoral politics, which makes it easier to compare re-election voting totals of legislators throughout the country.
Operationalisation
The notion of “electoral success” is pivotal and will be defined as the raw percentage of the popular vote that incumbent Senators seeking re-election received in the elections of 2012, 2014, 2016 and 2018. The main reason for choosing the Senate rather than the House of Representatives is that data on certain variables, such as GDP per capita and presidential approval ratings, are not easily available on a district-level.
A total of 30 incumbent Senators ran for re-election in 2018, resulting in 25 victories. In 2016, 29 Senators ran for re-election and 27 of them won; in 2014, 21 out of 26 Senators running again won their race. In 2012, 20 out of 21 were successful. Several cases, however, will be removed from the analysis. As we are comparing the number of missed votes throughout an entire term, we will not include incumbents that were elected through special elections and hence did not serve full six-year terms before running for re-election. Therefore, five cases (John Barrasso, Joe Manchin, Kirsten Gillibrand, Roger Wicker and Scott Brown) are removed
9
from 2012, three cases (Chris Coons, Ed Markey and Cory Booker) are removed from 2014, three cases (James Lankford, Brian Schatz and Tim Scott) are removed from 2016 and zero are removed from 2018. Furthermore, candidates without an opponent running on the ticket of a major party will be removed. This concerns Jeff Sessions in 2014, as Democrat Victor Sanchez Williams ran only as a write-in candidate. Due to California’s “jungle primary” system, Dianne Feinstein ran against another Democrat (Kevin de Léon) in 2018. Since many of the variables denote differences between Democratic and Republican candidates, Feinstein’s 2018 re-election effort will be removed (as opposed to her 2012 re-re-election, when she ran against a Republican). For similar reasons, Lisa Murkowski's 2016 re-election is not included, as her main opponent was a Libertarian. In total, 14 cases are removed and 92 remain.
The main independent variable is the number of votes that members of the Senate missed from the beginning of their term up until the date of their next election. These numbers are provided by GovTrack, a website tracking the legislative activities of members of the U.S. Congress (see ‘List of Representatives and Senators’ at https://www.govtrack.us/congress/members/current). Senators running for re-election in 2014 started their term in January 2009; those running for re-election in 2016 started their term in January 2011 and the last category consists of incumbents who started in January 2013 and ran for re-election in 2018. Not all Senators took up the same number of votes, but the numbers did not deviate much. Senators serving between 2007-2013 and 2009-2015 were both eligible for 1839 votes. Those serving between 2011-2017 could participate in 1645 votes, while Senators serving between 2013-2019 had the ability to vote 1758 times.
Deriving from the hypotheses, GDP per capita, unemployment numbers, ideological affinity, fundraising numbers and presidential approval are considered as well. These are measured as follows:
GDP per capita (H1): The Bureau of Economic Analysis collects annual data on the Real Gross Domestic Product (GDP) per capita of individual states in the United States (see ‘Real GDP in chained dollars’ at https://apps.bea.gov/iTable/). The real GDP per capita is measured in chained dollars, allowing for comparisons between different years. For every Senator, the percentage increase or decrease of GDP per capita their state, during their full term, is listed. For example, Senator Lindsey Graham from South Carolina ran for re-election in 2014, after starting his term in 2009. Between 2009 and 2014, the real GDP per capita in South Carolina rose from $36,352 to $38,053, or 4.68%.
10
Unemployment (H2): The Bureau of Labor Statistics collects data on the monthly unemployment rates for every state (see 'Unemployment rates, seasonally adjusted' at
https://www.bls.gov/lau/). For the analysis, the unemployment rate in the first month of an incumbent’s term is compared to the unemployment rate in their state on the month of the election. For example, Senator Angus King from Maine ran for re-election in November 2018, after starting his term in January 2013. From January 2013 to November 2018, the unemployment rate in Maine decreased from 7.1 to 3.5 percent, resulting in a -3.6 score in the analysis.
Party identification (H3): The variable of party identification is measured based on data by Gallup, an American polling organisation. The data used derives from polls by Gallup in 2012 (Gallup, 2012, see https://news.gallup.com/poll/160175/blue-states-outnumber-red-states. aspx), 2014 (Gallup, 2014, see https://news.gallup.com/poll/181475/massachusetts-maryland-democratic-states.aspx), 2016 (Gallup, 2016, see https://news.gallup.com/poll/ 203117/gop-maintains-edge-state-party-affiliation-2016.aspx) and 2018 (Gallup, 2018, see
https://news.gallup.com/poll/247025/democratic-states-exceed-republican-states-four-2018.as px) asking voters in each state whether they identify with the Democratic or Republican Party. The results are based on telephone interviews throughout the respective years, with random samples of 165,427 adults (2012), 177,034 adults (2014), 177,990 adults (2016) and ‘more than 73,000 adults’ (2018). The 2018 poll found that Maryland is the most 'Democratic' state, with 52.8 percent of respondents saying they identify with the Democratic party; Utah is considered the most ‘Republican’ state at 59 percent.
Ideological compatibility (H4): Ideological affinity between voters and their representatives is measured by combining two different scales. Firstly, the Cook Partisan Voting Index (CPVI) is used as an indicator of how strongly a state leans towards one of the major parties, compared to the nation as a whole. For example, a score of R+15 means that in the previous two presidential elections, that state voted an average of fifteen points more Republican than the nation did as a whole. Previous rankings are not stored on Cook’s website, but can be accessed through old versions of the Wikipedia article (see https://en.wikipedia.org/w/index. php?title=Cook_Partisan_Voting_Index&action=history). Secondly, GovTrack's ideology score of individual members of Congress will be used to measure to what extent the ideology of voters is reflected by their members of Congress. The ideology score is based on positions taken in roll-call voting, can range from 1 (very liberal) to 100 (very conservative) and has an average of 45 points for all Senators in for example 2014 (the average varies slightly for every
11
Congress). For each member, a score will be calculated to show how far they stray from their constituents. The state of New York had a CPVI score of D+11 in 2014. Hypothetically, a Senator from New York might be expected to be 11 points more liberal than the average Senator, and therefore be at 34 (45 minus 11) in GovTrack’s score. However, for example, Senator Chuck Schumer (D-New York) has an ideology score of 9 on GovTrack. The difference between his expected ideology score and his actual score amounts to 25, which will be used as the score for his ideological compatibility. A score of 0 indicates a perfect match. GovTrack creates individual report cards for every Congress, which can be found at
https://www.govtrack.us/congress/members (see ‘Report Cards’).
Assessment of the executive branch (H5): Polls by Gallup from 2012 (Gallup, 2012, see https://news.gallup.com/poll/160133/Obama-gets-highest-2012-job-approval-hawaii.aspx), 2014 (Gallup, 2014, see https://news.gallup.com/poll/181463/majorities-five-states-approved-obama-2014.aspx), 2016 (Gallup, 2016, see https://news.gallup.com/poll/189002/obama-rated-best-hawaii-2016-worst-west-virginia.aspx) and 2018 (Gallup, 2018, see
https://news.gallup.com/poll/247004/trump-job-approval-higher-states-2018.aspxs) are used to measure the approval ratings for President Obama (2012, 2014 and 2016) and President Trump (2018) in each state. These numbers constitute an annual average of Gallup’s daily tracking poll. Therefore, the numbers might have shifted throughout the year as the election came closer. However, no pollster provides more frequently attained state-by-state numbers for every election year considered. For Republican incumbents, the inverse of the approval rating for President Obama will be used in 2012, 2014 and 2016. For example, as 39 percent of adults in Tennessee approved of Obama in 2012, Republican Senator Bob Corker receives a score of 61 (100-39) in the analysis. The same applies for Democratic incumbents in 2018, during President Trump’s administration.
Campaign fundraising (H6): Fundraising totals gathered by the Federal Election Committee (FEC, 2019, see https://www.fec.gov/data/browse-data/?tab=candidates) are used to measure the financial prowess of incumbents and their challengers. The FEC requires candidates to file reports on their campaign fundraising totals. However, as campaigning in some states is inherently more expensive than in others (Bonneau, 2007), the variable will not consist simply of the amount of dollars raised in the campaign. Rather, an incumbent’s fundraising total will be presented as a percentage of the combined fundraising of both candidates. For example, Senator Elizabeth Warren raised $35,204,951.34 during the Massachusetts Senate race of 2018, while her opponent Geoff Diehl raised $3,025,033.56. Warren’s share amounts to 92.09% of
12
the total fundraising pool. One candidate, Mark Clayton (2012), did not file reports with the FEC. According to The Washington Post, which pondered if he was “America’s worst candidate”, he had raised a mere $278, which is the number used in this analysis.
Number of bills introduced (H7): The official website of the U.S. Congress stores data on the legislative activities of Senators (U.S. Congress, 2019, see https://bit.ly/38NB87g). Bills introduced include resolutions, concurrent resolutions and joint resolutions. As mentioned in the literature review, data on the number of bills from each Senator that came out of committee, as well as the number of bills from every Senator that actually became law are added as control variables. All of those data are accessible through the website of Congress, by using the filters on the left side of the screen. Furthermore, seniority will be measured by the number of full terms an incumbent Senator has served prior to their re-election effort.
Table 1. Frequencies of the dependent variable and independent variables
_____________________________________________________________________ N Min. Max. Mean St. deviation _____________________________________________________________________
Percentage of the vote 92 39.4 78.4 57.3 8.00
Missed votes 92 0 341 46 54.9 Seniority 92 1 7 1.95 1.37 GDP difference 92 -12.5 15 4.06 5.06 Employment difference 92 -5.4 6 2.05 2.82 Party identification 92 30 62 46.35 6.00 Ideological compatibility 92 0 47 22.32 9.62
Assessment ex. branch 92 32 81 55.75 10.48
Fundraising 92 24.10 100 76.58 19.72
Bills introduced 92 17 399 132.54 76.35
Bills out of committee 92 1 92 21.37 15.60
Bills that became law 92 0 15 2.85 2.76
_____________________________________________________________________
Table 1 displays the observed frequencies of all variables. On average, incumbent Senators received 57.3 percent of the vote in their elections. No Senator received less than 39.4 percent of the vote (Mike Pryor in 2014), while John Hoeven from North Dakota received the biggest vote share (78.4 percent in 2016). The observed average of 57.3 percent reveals that incumbents are generally successful in their re-election efforts. Two Senators, Susan Collins of Maine and
13
Chuck Grassley of Iowa, maintained perfect attendance records, not missing a single vote in the observed sessions of Congress. Mark Kirk of Illinois was absent most often, missing 341 votes. On average, Senators missed 46 votes during their terms. Another notable result is the relative financial prowess of incumbent Senators over their opponents. On average, Senators dominate more than three-quarters of the fundraising pool. Garnering sufficient funds to challenge an incumbent is an obstacle for non-incumbents. In just 11 of 92 cases, an incumbent raised less money than his or her opponent. Furthermore, regarding legislative activities, noticeably few bills ever become law in the United States. The average Senator introduced around 133 bills during their six-year term, but saw just around 3 of them being adopted. Most bills ‘die’ before they are considered by the relevant committee; if they are considered by the committee, bills still need to be brought to a floor vote, be adopted by both chambers and survive a potential presidential veto before being adopted as law.
Results
To analyse the data, an Ordinary Least Squares (OLS) regression model is used. The OLS regression model allows for the introduction of multiple variables simultaneously and provides coefficient estimates that display their effect on the dependent variable, which is the percentage of the vote received by incumbents. The assumptions of multiple regression analysis do not appear to be violated. Field (2009) prescribes that in a normally distributed sample, standardised residuals greater than 3.29 are cause for concern, as is having more than one percent of cases with a value greater than 2.58 or having more than five percent with a value greater than 1.96 (Field, 2009, p. 216). The only potential cause for concern is that one case, or 1.09 percent of the entire sample, has a value greater than 2.58 (that value being 3.11). Field prescribes using the value of Cook’s distance to measure the extent to which a specific case can bias a dataset; Cook and Weisberg (1982) have suggested that values greater than 1 may be a cause of concern. This specific case has a Cook’s distance value of just 0.07, suggesting a limited influence on the dataset. Furthermore, the assumption of homoscedasticity is not violated. This assumption states that the so-called statistical 'noise' in the relationship between the independent variables and the dependent variable is the same across all values of the independent variables. The statistical variances, therefore, are equal (for guidelines, see Field, 2009, pp. 149-152). Lastly, the indicators to detect multicollinearity, which is a highly linear relationship between two or more explanatory variables, do not raise red flags. VIF-values should not be greater than 5 (Field, 2009, pp. 224-225), while tolerance values should be above 0.2. Every variable in the dataset meets these assumptions. Table 2 displays the results of the
14
linear regression analysis. Two models are shown: Model 1 solely measures the effect of missing Senate votes on the vote share in the subsequent election, and Model 2 adds all the other variables mentioned in the part on the operationalisation.
Table 2. Linear regression model of vote share received by incumbents __________________________________________________ Model 1 Model 2 __________________________________________________ (Constant) 59.08 *** 13.51** (1.05) (3.90) Missed votes -0.04* -0.01 (0.02) (0.01) Seniority 0.15 (0.37) GDP difference 0.01 (0.11) Employment difference -0.46* (0.21) Party identification 0.44** (0.13) Ideological compatibility 0.03 (0.05)
Assessment ex. branch 0.21**
(0.07)
Fundraising 0.16***
(0.03)
Bills introduced 0.00
(0.01)
Bills out of committee -0.01
(0.04)
Bills that became law -0.04
(0.20) __________________________________________________ R2 0.07 0.75 Adj. R2 0.06 0.71 N 92 92 __________________________________________________
Note: OLS regression coefficients with standard error between brackets.
*** p <0.01, ** p <0.01, * p <0.05
As can be seen in Model 1 and 2, the amount of missed votes does not have a statistically significant effect on the vote share received by incumbents, when controlling for other factors.
15
The results show too much disparity to determine any positive or negative effect on the vote share incumbent Senators receive. On the one hand, voters re-elected the two Senators who did not miss any votes by resounding margins. Susan Collins of Maine and Chuck Grassley of Iowa received respective totals of 68.5 percent (2014) and 60.1 percent (2016) of the vote. In addition, Mark Kirk of Illinois missed the highest number of votes, 341, and he lost his Senate seat after receiving just 39.8 percent of the vote in 2016. On the other hand, Senators like Jim Inhofe of Oklahoma (68.0 percent in 2014) and Bernie Sanders of Vermont (67.4 percent in 2018) were not punished for missing relatively many votes. Knowing only the total number of missed votes, therefore, does not provide an indication of the vote share an incumbent will receive in their next election. The main hypothesis (H8) is rejected: the results do not indicate that missing votes leads to a lower vote share.
In fact, the indicators used to measure plenary activity do not appear to hold a relationship with electoral success, as the number of bills introduced, bills considered on the floor and bills eventually passed also do not show significant results. The three Senators that introduced the lowest amount of bills, Bob Corker of Tennessee, Jim Risch of Idaho and Richard Shelby of Alabama, all received at least 64 percent of the vote. Their three colleagues who introduced the highest number of bills, Dianne Feinstein of California, Bob Menendez of New Jersey and Bob Casey Jr. of Pennsylvania, were re-elected as well, albeit with a lower vote share. In between those ends of the spectrum, again too much disparity exists to determine any positive or negative effect. Therefore, H7 is rejected: the results do not indicate that introducing more bills leads to a higher vote share. A closer look at the other variables regarding legislative activity provides similar results. For example, out of the fifteen Senators who did not introduce a single bill that became law, just two received less than fifty percent of the vote in the subsequent election. These results do not indicate that voters do not take plenary activity into account when voting. Rather, they indicate that it is not enough to know only the sheer number of bills introduced, bills considered on the floor and bills passed into law.
Furthermore, the analysis does not provide support for H1 and H2, the variables regarding the economic conditions. The results do not indicate that an increase in GDP per capita in a state (H1), or a decline in unemployment in a state (H2), during an incumbent’s term, leads to a higher vote share in their next election. On the contrary, the results suggest that a declining employment rate (H2) slightly hurts incumbents (b = -0.46, p < 0.05). The value used in the analysis was the difference in the employment rate observed between the first and last year of an incumbent’s term. According to these results, the incumbent may expect an extra percent of
16
the vote share for every percent increase in the unemployment rate. A possible explanation for the peculiar outcome is that some states had a low unemployment rate to begin with, leaving little room to decrease it further. For example, North Dakota saw a decline of the unemployment rate from 3.5 to 2.9 percent during John Hoeven’s term between January 2011 and November 2016. That decline is clearly lower than the observed average of 2.05 (Table 1). Yet, voters re-elected Hoeven with 78.4 percent of the vote. In this case, unemployment was relatively low to begin with and voters therefore did not have reason to punish Hoeven. The results could, however, also indicate that a rise in unemployment leads voters to ‘stick with what they know’, rather than taking a risk by electing non-incumbents. Out of the 18 incumbents presiding over a declining employment rate, 16 received more than 50 percent of the vote in their next election. Their average vote share amounts to 58.7 percent, which is slightly higher but comparable to the overall average of 57.3 percent.
While H4 is not supported by the analysis, the results do not suggest that ideological compatibility between voters and Senators is unimportant to voters. They could suggest that voters do not take into account the entirety of an incumbent’s voting record, but rather votes on specific issues important to them. Also, due to the operationalisation of the data, incumbents close to the ‘middle’ are generally viewed as more compatible with voters in their state, especially in states with no substantial Republican or Democratic bias in Cook’s CPVI score. An incumbent with a centrist voting record from a state without a big bias towards one of the major parties receives a low ideology score, indicating a high level of ideological compatibility with the voters. An example is Bill Nelson from Florida, a state with just a slight Republican bias (R+2) in 2018. At 0.44 in GovTrack’s ideology score, Nelson displayed a centrist voting record theoretically in line with his constituents’ preferences. Nevertheless, he lost his seat by receiving just shy of 50 percent of the vote in 2018. Incumbents from swing states like Florida often confront tough re-election battles, as the opponent’s party continuously holds a substantial support base in the state. Incumbents in non-swing states might not appeal to the entirety of their party base, but due to the two-party system they will continuously attract support from a large part of their party-base, as the only viable alternative is the challenger from the opposition party.
The results do provide support for H3 and H5, which respectively concern voters’ party identification and their assessment of the executive branch. Party identification (H3) appears to be the strongest indicator of an incumbent’s vote share. For every additional percent of voters identifying with an incumbent’s party in a state, the incumbent may expect an extra 0.44
17
percent of the vote share. Voters, therefore, may be seen as likely to vote along party-lines in elections that concern their states’ representatives to the Senate. If they identify as Republicans, they are more likely to support a Republican incumbent; if they identify as Democrats, they are more likely to support a Democratic incumbent in the Senate. H5 concerns the assessment of the executive branch and has a b value of 0.21, indicating an increase of 0.21 percent of the vote share for every percent increase on the scale of presidential approval. If voters in a state approve of the president, they are more likely to support an incumbent Senator from their state of the same party; if they disapprove, they are more likely to support an incumbent Senator of the opposition party.
Lastly, the percentage of fundraising (H6) shows a statistically significant, expected increase of 0.16 percent of the vote for every extra percent of the ‘fundraising pool’ obtained by an incumbent. Zero out of the twenty Senators with the lowest share of the fundraising pool, ranging from 24.1 percent (Bob Menendez from New Jersey) to 60.5 percent (Joe Manchin of West Virginia), received more than 54 percent of the vote, while the average (Table 1) for all incumbent Senators is 57.3 percent. “America’s worst candidate” Mark Clayton, who raised $278 in 2012 according to the Washington Post, received 30.4 percent of the vote in Tennessee's 2012 Senate race, compared to 64.9 percent for his opponent Bob Corker, who controlled a rounded up share of 100 percent of the fundraising pool. These results highlight the importance of raising enough money for a political campaign. As mentioned above, the overwhelming majority of incumbents manage to raise more funds than their opponents, thereby solidifying their incumbency advantage.
Conclusion
The main goal of this research paper was to discern the relevant factors in determining the vote share incumbents receive. Based on previous studies, the influence of economic conditions, party identification, ideological compatibility between voters and incumbents, assessments of the executive branch, fundraising totals and plenary activity was measured in a case study of U.S. Senators. Consistent with existing literature (Samuels, 2001), voters are more likely to support well-financed campaigns, as well as incumbents sharing their party identification (Bartels, 1992). Also, voters approving of the executive branch are more likely to support incumbent Senators from the same party, and the opposite holds true for voters who disapprove of the executive branch (Hinckley, 1970; Key, 1958; Stewart, 1987).
18
The other hypotheses did not find support, including hypotheses regarding economic voting. These results warrant further research, as a different operationalisation may yield different results. In this analysis, voters were assumed to judge incumbents on the entirety of their term in office. It is conceivable, however, that voters are more concerned with the state of the economy on the day of voting, rather than the way it has improved or worsened in the past (six) years. Future researchers could consider using the change in GDP and/or unemployment rates for a shorter period, closer to the election, or merely use the observed rates on election day. Other indicators of the economy can be added, like the performance of the stock market, average household income or the consumer price index.
Missing votes, the subject of the main hypothesis, does not appear to hurt incumbents’ electoral prospects. Voters do not seem eager to punish politicians for missing votes, despite some observed historical cases of challengers successfully pointing out an incumbent’s lacklustre attendance record. When it comes to attendance records, voters appear to accept the ‘home style’ of representatives observed by Fenno (1978). In terms of attendance, the ‘chain of delegation’ mentioned by Strøm (1999) appears to be functioning well, as there is no indication of dissatisfaction from voters with their representatives’ attendance record. They do retain the ability to do punish representatives, which is an important tool at their disposal in terms of ensuring democratic accountability.
Additional research can help to shine a light on the electoral effects of vote-missing. For example, it is conceivable that incumbents are only hurt by their subpar attendance if challengers actively choose to attack them on it. Some challengers may not want to do this, as they might choose to focus on their own plans instead, or they might be vulnerable to similar counterattacks for their attendance record in other bodies that they previously served in. Much remains to be learned about the relationship between voters and representatives in a representative democracy.
References
Abramowitz, A. (1980). A Comparison of Voting for U.S. Senator and Representative in 1978. The American Political Science Review, 74(3), 633-640.
Abramowitz, A. (1988). Explaining Senate Election Outcomes. American Political Science
19
Abramowitz, A., & Segal, J. (1992). Senate Elections. Ann Arbor: University of Michigan Press.
Akirav, O. (2018). Electoral Rules and Legislators’ Productivity. Parliamentary Affairs,
72(3), 638-663.
Alesina, A., & Rosenthal, H. (1995). Partisan politics, divided government, and the
economy). New York: Cambridge University Press.
Ashworth, S. (2005). Reputational Dynamics and Political Careers. The Journal of Law,
Economics, and Organization, 21(2), 441-466.
Ashworth, S., & Bueno de Mesquita, E. (2008). Electoral Selection, Strategic Challenger Entry, and the Incumbency Advantage. The Journal of Politics, 70(4), 1006-1025. Bartels, L. (2000). Partisanship and Voting Behavior, 1952-1996. American Journal of
Political Science, 44(1), 35-50.
Benton, A. (2005). Dissatisfied Democrats or Retrospective Voters?: Economic Hardship, Political Institutions, and Voting Behavior in Latin America. Comparative Political
Studies, 38(4), 417-442.
Bonneau, C. (2007). The Effects of Campaign Spending in State Supreme Court Elections.
Political Research Quarterly, 60(3), 489-499.
Bowler, S. (2010). Private Members' Bills in the UK Parliament: Is There an 'Electoral Connection'? The Journal of Legislative Studies, 16(4), 476-494.
Brown, A., & Goodliffe, J. (2017). Why do Legislators Skip Votes? Position Taking Versus Policy Influence. Political Behavior, 39(2), 425-455.
Burden, B., & Wichowsky, A. (2014). Economic Discontent as a Mobilizer: Unemployment and Voter Turnout. The Journal of Politics, 76(4), 887-898.
Campbell, A. (1980). The American voter. Chicago, Ill: University of Chicago Press.
Campbell, J. (1997). The presidential pulse of congressional elections (2nd ed.). Lexington, Ky.: University Press of Kentucky.
Campbell, J., & Sumners, J. (1990). Presidential Coattails in Senate Elections. The American
20
Canes-Wrone, B., Brady, D., & Cogan, J. (2002). Out of Step, Out of Office: Electoral Accountability and House Members Voting. American Political Science Review,
96(1), 127-140.
Capella, L., & Taylor, R. (1992). An analysis of the effectiveness of negative political campaigning. Business and Public Affairs, 18(2), 10-17.
Çarkoğlu, A. (1997). Macro Economic Determinants of Electoral Support for Incumbents in Turkey, 1950–1995. New Perspectives on Turkey, 17, 75-96.
C-SPAN. (2012). [Video]. Retrieved November 12, 2019, from https://www.c-span.org/ video/?308797-1/florida-senate-debate
Cohen, L., & Noll, R. (1991). How to vote, whether to vote: Strategies for voting and abstaining on congressional roll calls. Political Behavior, 13(2), 97-127.
Cox, G., & Terry, W. (2008). Legislative Productivity in the 93d–105th Congresses.
Legislative Studies Quarterly, 33(4), 603-618.
Dahl, R. (1989). Democracy and its critics. New Haven, London: Yale University Press. Downs, A. (1957). An economic theory of democracy. New York: Harper & Row.
Ebert, J. (2018). Diane Black has most missed votes in 2018 among Tennessee congressional delegation. Retrieved November 12, 2019, from https://eu.tennessean.com/story/news/ politics/2018/03/13/diane-black-has-most-missed-votes-2018-among-tennessee-congressiona l-delegation/418506002/
Everdsen, A. (2018). Beto O’Rourke says Ted Cruz has an attendance problem. Cruz has missed 14 percent of votes; O’Rourke has missed 3 percent. Retrieved November 12, 2019, from https://www.texastribune.org/2018/10/16/here-are-ted-cruz-and-beto-orourkes-attendance-numbers/
Fearon, J. (1999). Electoral Accountability and the Control of Politicians: Selecting Good Types versus Sanctioning Poor Performance. In Democracy, Accountability, and
Representation (pp. 55-97). Cambridge: Cambridge University Press.
Fenno, Richard F. (1978). Home style: House Members in their districts. Boston: Little, Brown.
21
Ferejohn, J., & Fiorina, M. (1974). The Paradox of Not Voting: A Decision Theoretic Analysis. The American Political Science Review, 68(2), 525-536.
Field, A. (2009). Discovering statistics using SPSS (and sex and drugs and rock 'n' roll) (3rd ed.). Los Angeles, CA: Sage.
Gerring, J. (2004). What Is a Case Study and What Is It Good for? American Political
Science Review, 98(2), 341-354.
Glazer, A., Grofman, B., & Owen, G. (1998). A neo-Downsian model of group-oriented voting and racial backlash. Public Choice, 97(1), 23-34.
Hagen, L. (2016). Murphy agrees to half the number of debates Rubio proposed. Retrieved November 12, 2019, from https://thehill.com/blogs/ballot-box/senate-races/295567- murphy-agrees-to-half-the-number-of-debates-rubio-proposed
Hinckley, B. (1970). Incumbency and the Presidential Vote in Senate Elections: Defining Parameters of Subpresidential Voting. The American Political Science Review, 64(3), 836-842.
Jost, J. (2017). Ideological Asymmetries and the Essence of Political Psychology. Political
Psychology, 38(2), 167-208.
Kahn, K., & Kenney, P. (1997). A Model of Candidate Evaluations in Senate Elections: The Impact of Campaign Intensity. The Journal of Politics, 59(4), 1173-1205.
Kalt, J., & Zupan, M. (1984). Capture and ideology in the economic theory of politics.
American Economic Review, 74(3), 279-300.
Kau, J., & Rubin, P. (1979). Self-Interest, Ideology, and Logrolling in Congressional Voting.
The Journal of Law and Economics, 22(2), 365-384.
Key, V. O. (1958). Politics, parties, and pressure groups (3rd ed.). New York: Crowell. Key, V., & Cummings, M. (1966). The responsible electorate: Rationality in presidential
voting, 1936-1960. Cambridge, Mass.: Belknap Press of Harvard University Press.
Kramer, G. (1971). Short-Term Fluctuations in U. S. Voting Behavior, 1896-1964. The
American Political Science Review, 65(1), 131-143.
Krasno, J. S. (1997). Challengers, competition, and reelection: Comparing Senate and House
22
Levy, J. S. (2002). Qualitative methods in international relations. In M. Brechter, & F. P. Harvey (Eds.), Millennial reflections on international studies (pp.432-454). Ann Arbor: University of Michigan Press.
Lott, J., & Reed, R. (1989). Shirking and sorting in a political market with finite-lived politicians. Public Choice, 61(1), 75-96.
Mayhew, D. (1974). Congress: The electoral connection. New Haven, London: Yale University Press.
Mintz, E. (1996). Members' roll call attendance sets election-year record. Congressional
Quarterly Weekly Report, 55(50), 3439.
Nadeau, R., & Lewis-Beck, M. (2001). National Economic Voting in U.S. Presidential Elections. The Journal of Politics, 63(1), 159-181.
Nelson, J. (2012). Sebring says Moore is ‘most absent member’ of Wisconsin’s congressional delegation. Retrieved November 12, 2019, from https://www.politifact.com/wisconsin/st atements/2012/oct/08/dan-sebring/sebring-says-moore-most-absent-member-wisconsins-c/
Nordhaus, W. (1975). The political business cycle. Review of Economic Studies, 42(2), 169-190.
Norpoth, H. (1996). Presidents and the Prospective Voter. The Journal of Politics, 58(3), 776-792.
Payne, J. G., & Baukus, R. A. (1985). Trend analysis of the 1984 GOP senatorial spots.
Political Communication, (5)3, 161-177.
Persson, T., Roland, G., & Tabellini, G. (1997). Separation of Powers and Political Accountability. The Quarterly Journal of Economics, 112(4), 1163-1202.
Rosas, G., & Shomer, Y. (2008). Models of Nonresponse in Legislative Politics. Legislative
Studies Quarterly, 33(4), 573-601.
Salmore, S., & Salmore, B. (1989). Candidates, parties, and campaigns: Electoral politics in
America (2nd ed.). Washington, D.C.: CQ Press.
Samuels, D. (2001). Money, Elections, and Democracy in Brazil. Latin American Politics
23
Sherman, A. (2012). Karen Harrington says Debbie Wasserman Schultz "missed 62 congressional votes— one of the worst records of any member of Congress”. Retrieved November 12, 2019, from https://www.politifact.com/florida/statements/2012/jan/ 17/karen-harrington/harrington-says-wasserman-schultz-missed-62-congre/
South Dakota Democratic Party. (2018) Noem Neglects Representing SD in DC to Advance Next Step in Her Political Career. Retrieved November 12, 2019, from
https://medium.com/@SDDP/noem-neglects-representing-sd-in-dc-to-advance-next-step-in-h er-political-career-332c6bc14255
Strøm, K. (2000). Delegation and accountability in parliamentary democracies. European
Journal of Political Research, 37(3), 261-290.
Westlye, M. C. (1991). Senate elections and campaign intensity. Baltimore, Md: Johns Hopkins University Press.
Woon, J. (2012). Democratic Accountability and Retrospective Voting: A Laboratory Experiment. American Journal of Political Science, 56(4), 913-930.
Wright, G., & Berkman, M. (1986). Candidates and Policy in United States Senate Elections.
The American Political Science Review, 80(2), 567-588.
Wright, M. (1993). Shirking and political support in the U.S. Senate, 1964–1984. Public
