Louisiana Sheriff Elections

Introduction “The state is considered the sole source of the ‘right’ to use violence. ” (Weber 1972, 1) County sheriffs in the United States are elected and empowered by their constituents to make arrests, conduct searches and seizures, and use deadly force in performance of their duties if public safety is at risk. Yet for all of this authority, sheriff elections have been largely ignored by political science. This study seeks to lay the foundation for further study into a position that is an amalgam of law enforcement officer and political entity. As mentioned previously, a sheriff is not just an elected cop.

Sheriffs are granted a variety of implied, reserved, and expressed powers; some of which are very similar to those granted to other political entities. Also like other political entities, federalism causes differences in powers and duties from one state to another such as in Louisiana where the sheriff runs the parish prison, performs law enforcement duties throughout the parish, and is the ex-officio tax collector and Pennsylvania where sheriffs only run the county jail and act as officers of the court.

This study focuses on sheriff elections in the state of Louisiana. Louisiana has a unique cultural history that includes being owned by the Spanish and French prior to being purchased by Norman Clark Louisiana Sheriff Elections 3 the United States. Because of its solid Roman Catholic beginnings, counties in Louisiana are referred to as parishes like a church parish. Numerous studies have found that policy responsiveness is a major factor in a legislative incumbent getting reelected (e. g. , Abramowitz 1988; Canes-Wrone, Brady, and Cogan 2002; Hogan 2008; Wright and Berkman 1986). The question arises, how to judge if an incumbent sheriff is performing in a satisfactory manner?

With a regular elected official, a voter could look at road construction, how our children are performing in schools, or if our taxes are going up or down. The primary policy goal of any law enforcement official is to lower crime rates. With this in mind, this study proposes using crime rate change between the year prior to and the year of the election to determine a sheriff’s policy responsiveness. V. O. Key, Jr. explored the concepts of political accountability and representation for incumbents (1961). Other studies show that voters refer to policy decisions made by incumbents (Hogan 2008; Maestas 2000).

Thanks to technological improvements it is far easier now for the average voter to access crime statistics either from the FBI or their local sheriff’s offices. The ‘Know-nothing’ studies championed the idea that the average voter does not have the knowledge to make informed decisions in the voting booth (Bennett 1988, 1996; Hyman and Sheatsley 1947; Somin 2004). The crime statistics can now even be taken down to the neighborhood level (Lee and Shihadeh 2009; Levit 1997; Shihadeh 2009). Because of the availability of crime information, citizens should have a better understanding of what crime rates are and make their votes accordingly.

Candidate quality also plays an important part in legislative elections. Van Dunk defined challenger quality as a candidate who has previously held public office (1997). Van Dunk makes an excellent point. Experienced politicians have made the connections that make it easier to raise campaign contributions as well as gain endorsements. Those traits would aid someone running for any elected office. Being a law enforcement office, it is conceivable that law enforcement Norman Clark Louisiana Sheriff Elections 4 experience would also be a quality that voters would desire in a sheriff.

For this study candidate quality will be both political experience and law enforcement experience. Many looking at the history of Louisiana sheriff elections would come to the conclusion that the residents of the state are racist and sexist due to the fact that only three women and fewer than five minorities have held the office of sheriff in the state. Gender and minority studies into elections have shown that women who are mothers are less likely to enter politics due to their children and minorities are more likely to come out in larger numbers when a quality minority candidate enters a race (Barreto 2005; Barreto, Villareal, and Woods 2005; Fox 2010;

Fox and Lawless 2011; Leal, Barreto, Lee, and de la Garza 2005). Keeping this in mind, few women are willing to run for an office that requires them to put their lives on the line opening the possibility that their children will be left without a mother. An examination of racial inequality in Louisiana would fill several bookshelves so this study will simply state that it does exist. This study looks to fill several gaps in knowledge that exists in sheriff elections.

The position carries a great deal of power and authority as well as a number of factors that could affect who voters chose to fill the office. Any results will provide for future studies into the subject with the possibility of expanding the study into other local elected law enforcement officers including city constables, city marshals, police chiefs of small departments, and county constables. Previous Studies As mentioned above studies into American sheriff elections are almost non-existent leaving this study to rely heavily on studies into normal political elections. The studies will include some that cover the areas of incumbency, candidate quality, retrospective voting, and crime.


The lone study into American sheriff elections was conducted by Raymond Surette, who examined the 1976 Florida county sheriff elections (1985). Surette compiled a data set that Norman Clark Louisiana Sheriff Elections 5 included political data, crime statistics, and socioeconomic data. This study will mimic several of the variables and approaches made by Surette, with the primary differences being that this study will examine more than one election cycle and will be examining Louisiana sheriff elections.

The state of Florida has 67 counties. In the 1976 elections, thirteen incumbents faced no opposition and were removed from the data set, leaving fifty-four counties in the study. This is one of the few short comings of Surette’s study and one that this study attempts to improve upon by examining multiple cycles. Were Surette’s findings only valid for that year, was it an outlier, or were the findings a trend? Surette found that as murder rates in a county went up, the incumbent’s reelection rates went up. However, if the number of rapes increased the incumbent could expect to lose votes. This study uses both of those crimes along with interactive variables for law enforcement and incumbency with crime.

Surette makes the statement that murders “seemed to either tap or result in a need for continuity on the part of the electorate, while rape added to the pressures for change” (Surette 1985, 24). This assumption reinforces the need for more study into this field; if a better understanding of the psyche of a crime victim can be developed the law enforcement agencies tasked with crime prevention would be able to respond with increased efficiency. If voters react to crime at the voting booth, the way that a sheriff or deputies handle crime is of importance. Responding to a crime involves more than just getting a deputy or the sheriff to a crime scene in an expeditious manner, it also involves the time frame that is needed to investigate the crime.

Often victims of domestic violence or sexual assaults are reluctant to come forth and report the crime (Dartnall and Jewkes 2013). Know Nothings Key and Cummings argue that voters used past experience to make decisions on voting (1966). The argument is made that citizens are not attempting to forecast the future when they vote, rather they simply use the candidate’s past performance to predict what the candidate will Norman Clark Louisiana Sheriff Elections 6 do in the future.

In this study, it is of interest to determine if voters use crime rates as their basis for their retrospective voting. If crime rates increase are incumbents held responsible? Are candidates with law enforcement experience rewarded at the polls? Voter knowledge is an important aspect of democracy. Thomas Jefferson is often quoted as saying, “An informed citizenry is the bulwark of a democracy. ” Regardless of if Jefferson actually made the statement it shows the importance of the statement that it is attributed to him. Many studies have shown that the average voter does not have a high political IQ.

Somin points out that an uninformed voter can be swayed by the elite and place doubts that democracy best serves the interests of the people when their votes are made randomly or with the latest television campaign advertisements (Somin 2004). Bennett followed up on the Hyman and Sheatsley (1947) study exploring the ‘know-nothings,’ attempting to assess if the new age of television led to a passive learning that increased voter knowledge (1988). Bennett found that only thirty percent of adults could correctly answer questions about politics, raising doubts that the average citizen understands the concept of U.S. democracy and why the country should continue using its brand of democracy.

The ‘know-nothings’ concept is of special interest in this study as local sheriff elections are among the purist forms of democracy, allowing voters to hold their local candidates responsible for their job performance at the lowest level of federalism. Romer, Jamieson, and Aday (2003) used a model containing five years of General Social Surveys and a survey of over 2,300 citizens of the city of Philadelphia to explore their idea that individuals who watched their local television news had an increased fear of becoming crime victims themselves.

The study notes that local news media depend on the sensationalism inherent in mayhem and violence, placing an emphasis on homicide and other violence. The study notes that due to the coverage of violent crime, citizens’ expectations of being the victims of crime increase along with a presumed rise in emotions of helplessness of being able to do anything Norman Clark Louisiana Sheriff Elections 7 about the crime rates. Romer, Jamieson, and Aday call their theory “The Television Exposure Hypothesis,” and is used here to support the hypothesis that voters are more aware of crime rates which allow them to make more informed decisions at the voting booth.

Candidate Quality Studies on candidate quality can be drawn from a number of different academic disciplines. The first looked at here is from Donald Wittman of the UC Santa Cruz Department of Economics (2007). Wittman explains that when voters have a low political IQ they will pay attention to pressure groups and endorsements. It is also noted that these groups will strive to tell the truth about the candidates. This study does not include a variable for endorsements, but includes the Wittman paper to show where voters can obtain information on candidate quality.

Namely from different pressure groups, endorsement groups, and technology. Candidate quality studies have examined candidates in both open seat races and those in which the incumbent is seeking reelection (Bianco 1984; Bond, Covington, and Fleisher 1985, 1997; Cannon 1990; Hall and Bonneau 2006; Jacobson 1990, 1992). This study will include candidate quality in both open elections and those with incumbents. Politically experienced candidates are considered to be quality candidates. Some studies have found that politically experienced candidates normally run in open elections (Bianco 1984; Canon 1990).

This can partially be explained by the incumbency advantage theory that shows the percentage of incumbents that get reelected in different elections. Experienced candidates would know how hard it is to beat an incumbent and avoid races with incumbents. Studies into judicial elections provide relevant information for this study (Bond, Covingston, and Fleisher 1985; Hall 1995, 2001a, 2001b; Hall and Bonneau 2006). Hall and Bonneau found that in 208 elections from 1990 through 2000 experienced candidates were able to somewhat offset the incumbent advantage. The study found that experienced candidates received a higher vote total than the candidates without experience.

Norman Clark Louisiana Sheriff Elections 8 Hypotheses The following are the hypothesis tested in this study, and are theorized to have a large impact on a sheriff elections: H0 = Having law enforcement experience (X4) or political experience (X5) has no impact on the share of the vote a candidate receives. H1 = Voters do not hold incumbents accountable for increases or decreases in crime rates.

Sample Selection Louisiana was selected as the focal point of this study due to its use of a jungle primary system that is not used in any other state in the South, and the high level of crime the state has had for the past two decades. The Louisiana electoral system has been called an open primary, a jungle primary, and a Cajun primary. Regardless of the name, the system is used in elections every four years in the year prior to presidential elections.

The system forces all of the candidates in an election to face each other regardless of political party affiliation. If a candidate receives over fifty percent of the vote in the primary that candidate wins the election outright. If no candidate received a majority vote, then the two candidates with the highest vote total will face each other in a runoff election one month later. Louisiana was selected as the focal point of this study due to its use of a jungle primary system that is not used in many other states and the high level of crime the state has. This study begins the research into Louisiana sheriff offices, laying the groundwork for further studies that examine other states.

The Louisiana electoral system has been called an open primary, a jungle primary, and a Cajun primary. The jungle primary is held every four years in the year prior to a presidential election. The system forces all the candidates for an office to face each other, regardless of political party affiliation. If a candidate receives over fifty percent of the vote in an election, that candidate wins the election outright. However, if no candidate received a majority vote then the two candidates that received most votes will face each other in a runoff election a month later. Norman Clark Louisiana Sheriff Elections 9 This system often results in candidates of the same political party competing against one another in the runoff (Jones 1997).

The second reason for choosing Louisiana is because of its crime rates. According to US.

News and World Reports, the state of Louisiana has ranked as the most dangerous state in the country for the last twenty years. 1 Louisiana also ranks among the top ten states for high poverty rates, high incarceration rates, and low education rates. These factors together emphasize the need for a sheriff to be a politician as well as a law enforcement officer, using the political ties of the office to improve education. Data, Methodology, and Variables The data for this study comes from four main sources: the FBI’s Uniform Crime Reports; the United States Department of Labor; the Louisiana Secretary of State’s Official Election.

Results; and the Louisiana Sheriffs Association. Dozens of other websites were accessed in order to gather information on candidate quality from campaign websites, newspapers, and the websites for the various sheriff’s offices used in the study. Data from the above locations was used to explore the six Louisiana sheriff election cycles that took place from 1987 through 2007. The variables for crime rates were taken from the year prior to the elections and the year of the elections. Uniform Crime Reports The Federal Bureau of Investigation (FBI).

Collects crime data from over 18,000 law enforcement agencies throughout the United States each year in order to meet the need of those agencies to have a reliable way in which to track crime. Each agency uses the same forms when submitting their crime statistics, allowing the FBI to track crimes across jurisdictions even if the crimes are called different names such as theft and larceny. There are instances each year where an agency will not file their UCR reports due to changes in management, personnel shortages, or 1 U. S. NEWS AND WORLD REPORTS. LOUISIANA MOST VIOLENT STATE IN THE U. S. RETRIEVED ON 09/02/2012 FROM HTTP://WWW. USNEWS.

COM/NEWS/ARTICLES/2012/05/07/LOUISIANA-MOST-VIOLENT-STATE-IN-THE-US . UPDATED ON 09/02/2012. Norman Clark Louisiana Sheriff Elections 10 computer problems to name a few. 2 In the instances that the agency did not supply UCR reports for the months that this study examines the applicable parish election is not included in the model when the crime statistics are applied. News Media Multiple news media web sites were examined in order to obtain information on candidates such as their political or law enforcement experience. The various sites were also used to verify the terms that the incumbents served.

Due to the rural nature of some of the parishes involved in this study, not all of the candidate quality measures could be obtained for all candidates. Some of the candidates did not file campaign finance reports nor use any types of social media to announce their intention to run for office. In the cases where the information was unavailable, those candidates were removed from the study. Model Outline This study will use a linier regression model and frequency distribution to analyze the data.

The independent variables are separated into several categories and are explained in depth later in this study. The first category is simply the dependent variable, which is the percentage of the vote carried by each candidate in the election. The second category consists of the political variables which are made up of incumbency and candidate political party affiliation. The third category is candidate quality which includes law enforcement experience and political experience. The final category is that of crime variables. The crime variables are separated into murder, rape, robbery, and burglaries.

Assault and Battery were not included in this study due to the difference in what constitutes each crime in different jurisdictions. The Dependent Variable The dependent variable (Y) in this study is the percentage of the vote carried by each candidate in the election. Other studies limit the inclusion of candidates to those that carry more than 5 or 10 percent of the votes in their race. This study includes all candidates regardless of the 2 THE FEDERAL BUREAU OF INVESTIGATION. “WORD ABOUT UCR DATA. ” RETRIEVED ON 01/10/2013 FROM HTTP://WWW. FBI. GOV/ABOUT-US/CJIS/UCR/WORD . UPDATED ON 01/10/2013. Norman Clark Louisiana Sheriff Elections 11 portion of the vote that candidate received.

I expect that the vote percentage received by each candidate will be highly susceptible to the variables that list if the candidate did or did not have of political and/or law enforcement experience. Political Variables The first independent variables in the study is labeled X1, and is the candidate’s political party affiliation. The variable is a dummy variable and recognizes if the candidate is a Republican. It is coded 1 for Republican and 0 for all others. In Louisiana eighty percent of the candidates for local sheriff list Democrat for their political party affiliation even though the

Republican Party currently controls the Louisiana Senate and House of Representatives. The second political variable is labeled X2 and stands for Incumbency. It is also a dummy variable and is coded 1 for yes and 0 for no. The purpose of this variable is to determine if an incumbent sheriff in Louisiana enjoys the same incumbent advantage that state and federal legislators do. Incumbency reelection rates at both the state and federal level have been found to be over 90%. The third political variable is labeled X3 and is the number of candidates in the various sheriff elections.

It is presumed that the more candidates there are in an election the more the vote share will be split. The jungle primary in Louisiana will allow two candidates of the same political party to face each other in a runoff. Candidate Quality Variables The candidate quality variables address the experience or lack of the same that the candidate running for election has and are also dummy variables. The first candidate quality variable is labeled X4 and stands for law enforcement experience, meaning the candidate has law enforcement experience.

The second candidate quality variable is labeled X5 and stands for if the candidate has political experience. Both variables were coded 1 for yes and 0 for no. It is theorized that a candidate who has held public office will understand the nuances of running for public office and will have made acquaintances with other politicians and donors that can provide the candidate with funding, both of which can help the candidate get elected. A Norman Clark Louisiana Sheriff Elections 12 candidate with prior law enforcement experience can point to that experience as a qualification to be elected due to familiarity with police procedures and state laws, as well as familiarity with the types of crimes that exist in different areas of the jurisdiction and how to identify those crimes and their culprits.

Crime Variables The final category of independent variables included in this study is crime variables. They are determined by the uniform crime reports that each parish submitted to the FBI each month. The crime rate is calculated by dividing the population of the jurisdiction by 100,000 and then dividing the number of crimes in that jurisdiction by that number, providing us with the number of crimes per 100,000 people. The percentage change between the previous year and the election year is then used as the crime variable. The first of these variables is labeled X6 and stands for the percentage change in the murder rate for each parish involved in the study. The percentage change in the number of rapes per 100,000 people is labeled X7.

These variables are also interactive variables and are multiplied by the variable for Law Enforcement Experience labeled X4 and by the Incumbent variable X2. It is theorized that a candidate with law enforcement experience will receive a reward or punishment for this experience by the voters. The interactive variables are labeled X8 for Law Enforcement Experience times Murder rate (X4*X6); X9 for Law Enforcement Experience times Rape rate (X4*X7); X10 for Incumbent times Murder (X2*X6); and X11 for Incumbent times Rape (X2*X7).

Model Outline When the variables are combined the model of candidate vote percentage is revealed. The first model will be run four times. Per Brambor, Clark, and Golder (2006) the interactive variables are included in each of the models along with the initial variables. The model is formulated thusly: Y= a + b1X1 + b2X2 + b3X3 + b4X4 + b5X5 + b6X6 + b7X7 + b8(X4*X6) + b9(X4*X7) + b10(X2*X6) + b11(X2*X7).

Findings Norman Clark Louisiana Sheriff Elections 13 The purpose of this study was to determine if crime rates, candidate quality, and incumbency have an impact on the vote share candidates receive in Louisiana sheriff elections; it has been successful. The model was run seven times; once for each of the election cycles and once with all of the cycles added together, the master model. The purpose of the master model was to develop a baseline by which the other models could be compared. The correlation between the observed and predicted values of the percentage of the vote the candidate carried was very consistent throughout the seven model runs.

The lowest R value was for the 2003 cycle, which was . 781. The highest R value was found in the 1987 cycle, which was . 875. The R value for all the elections in the master model was . 811. The Adjusted R Square for each of the models was also fairly consistent. The 2003 cycle had an Adjusted R Square of . 566 and the master model had a . 651. A critical significance level of . 001 was used for all of the models. The closest critical value was also in the 2003 cycle where the F value was 14. 164 with a critical F value of 2. 35. Incumbency was significant at the . 001 level in every model.

This is exactly what was expected. The variable was coded 1 if the candidate was the incumbent. Reading the output of the total election model, incumbency is worth 25. 94 percent of the vote share. At the federal and state level incumbency has been found to be significant and appears that way in Louisiana sheriff elections as well. The variable for political party was not significant in any of the models. It should be noted that political party had a negative coefficient for a Republican candidate. This output was expected as most of the local level elected officials are Democrats in the state.

The no-brainer for this study was the variable for the number of candidates, which was significant in every model. The more candidates in a race the fewer votes each candidate can expect to carry. In the master model for every candidate in the race a candidate can expect to lose 3. 7% of the vote. Norman Clark Louisiana Sheriff Elections 14 In the master model the two candidate quality variables (law enforcement experience and political experience) were both significant. In the yearly models the two variables alternated being significant. This could be due to the lack of resources to properly verify a candidate’s

experience in any given election cycle. None of the crime variables were significant in any of the models. This was unexpected. It was theorized that the percentage change of rapes or murders would be significant. Additionally, none of the interactive variables were significant. There were some interesting results in the incumbent interaction variables in the master model. For every one unit increase in the murder variable the incumbent can expect a . 017 increase in the vote share for the incumbent. However, for every one unit increase in the rape variable the candidate can expect a . 015 decrease in the vote share.

1987 The 1987 election cycle is the earliest cycle of which the Louisiana Secretary of State has converted to digital form, allowing easy access for anyone in the country with access to a computer and the internet. There were some difficulties obtaining information about some candidates, which occurred in most of the cycles. The 1987 cycle provides several interesting findings. The first is that of the fifty incumbents who sought reelection, forty-six of them won their election for a ninety two percent winning percentage. The adjusted R square for the 1987 cycle was . 737, a very high output that.

advises there is a strong association in the model. The F Statistic outcome is 26. 938 while the critical F statistic for a . 01 level of 2. 45, leaving a positive difference of 24. 49. In 1987, a candidate who is the incumbent can expect a 34. 01 increase in vote percentage. The incumbent variable is significant at the . 001 level. Likewise, the variable for Number of Candidates is significant at the . 001 level. The output shows that for every candidate who enters the election, a candidate can expect to lose 3. 18 percent of the vote. Law Enforcement Experience was also significant at a .

05 level while giving the candidate with law enforcement Norman Clark Louisiana Sheriff Elections 15 experience approximately an 8. 602 boost in vote percentage. Interestingly enough, none of the crime rates was significant, however a candidate with law enforcement experience or the incumbent could expect an increase in their vote share for every percentage point change in crime rates. Insert Model 1 & 2 here. 1991 The 1991 election cycle produced an Adjusted R Square of . 568, which is a strong correlation. It should be noted that there were 11 variables in this model with 70 cases examined.

It is preferable to have more cases however, there are not always a high number of cases upon which to draw. The critical F score for a . 01 confidence level is a 2. 56 and the output in this case is a 9. 362 which more than meets the requirements. A candidate’s political party is not significant nor does it give a significant boost to the vote percentage. Incumbency in this model is highly significant and provides the incumbent candidate a boost of 27. 8 to his/her vote percentage. Although incumbents in 1991 did not fare as well as their 1987 counter parts in seeking reelection, they still had a win-loss record of 37-8 which is a winning percentage of 78. 38. In number of candidates in the race is stronger in this cycle than in the prior election costing 4.

472 for every candidate that enters the race. A candidate with law enforcement experience has a stronger boost in this cycle than before, seeing a boost of 14. 385 in vote share. In this cycle a candidate with political experience cannot expect anywhere as near a boost. As in the 1987 election, crime rate changes are not significant. 1995 The 1995 cycle had a stronger Adjusted R Square than in 1991, showing a correlation of . 615. Our number of cases increases as well to 121, giving us a critical F score of 2.

4 and an actual F score of 20. 14 which is highly significant. Incumbency was not as important in this election though it was still significant, but only provided the incumbent a boost of 19. 67 in vote Norman Clark Louisiana Sheriff Elections 16 share. Candidates could still expect to lose just over 4 points in vote share for every candidate that enters the race. 1995 was the first election in which a candidate with political experience could expect a larger boost to their vote score than someone with law enforcement experience. The question to be considered here is why the change occurred.

Both types of experience were statistically significant. Yet again, crime rate change was not significant. Insert Models 3 & 4 here. 1999 The adjusted R square score was again significant, showing a correlation of . 679. The F score for the year was 24. 224 which was far greater than the critical F score of 2. 41. Incumbency was again highly significant with a coefficient of 31. 725 and a win-loss ratio of 38-4. Experience was again important with political experience having a significance score of . 097 while being worth 6. 247 of the vote percentage.

Law enforcement was the strong of the two being significant at the . 02 level and providing a vote percentage boost of 8. 81. 2003 The adjusted R square for the 2003 election dropped down to a . 569, but still shows a fairly strong correlation. The F score in this election was a 15. 55 was smaller than the previous election but was still significant and larger than the critical