ECONOMETRICS: Discrimination in Peremptory Challenges
Share
Introduction
Peremptory challenges, which allow attorneys to remove jurors without explanation or cause during the voir dire process, have been limited by two Supreme Court cases, Batson v. Kentucky (476 U.S. 79 [1986]) and J.E.B. v. Alabama ex. rel. T.B. (511 U.S. 127 [1994]). These court cases have ruled it unconstitutional for attorneys to remove potential jurors based solely on race or gender. However, more recent literature has shown that attorneys continue to use demographic characteristics to remove potential jurors they believe are likely to favor the opposition (Anwar et al. 2010; Anwar et al. 2014; Flanagan 2018; Beck 1998). This is possibly because it is very easy for attorneys to get around these Supreme Court rulings by “camouflaging” their discriminatory removals with any other explanation for removal (Beck 1998).
My research is similar to Flanagan (2018), where data is used to determine whether or not discrimination exists in peremptory challenges and whether it’s strategic. However, my research will only analyze the first part, whether or not attorneys are using peremptory challenges to discriminate potential jurors based on race, gender, and age. Flanagan compares attorney strike patterns to the actual effects of a jury composition on the probability of conviction in order to see whether the discriminatory principles being followed by attorneys are based on false stereotypes or strategic and tied to actual group behaviors (2018). Schwartz and Schwartz (1996) also addresses this concept, arguing that the removal of potential jurors based on demographic information is only significant if relevant attitudes vary systematically between groups. This is because if attorneys are not simply discriminating based on false stereotypes, they are removing a group of people with particular attitudes (not necessarily bias) and will create a jury that is not representative of a population’s attitudes. This is important because a fair and impartial jury is protected under the Sixth Amendment. I believe a tool that allows attorneys to purposefully remove potential jurors based on something other than their inability to be impartial is solely strategic and does not aid in establishing a fair and representative system. Proof that attorneys are using peremptory challenges in this way should encourage reconsideration of the system in place.
Using data collected by the North Carolina Jury Sunshine Project (NCJSP) that observes potential jurors in North Carolina in 2011, I will present four linear probability models to assess an attorney’s likeliness to use a peremptory challenge in order to exclude a potential juror based on race, age, or gender. I will also include interaction variables to illustrate how the defendant’s race may have an effect on the magnitude of the effect of race.
The statistically significant coefficients of the regressions show that black potential jurors are removed disproportionately by the state, a juror’s age is negatively correlated with the probability of the state removing the potential juror, and the race of the defendant has a significant effect on the magnitude of the effect of race. If there is a black defendant, it is more likely that the state will strike a black juror. When it comes to the defense, they are much less likely to remove a black potential juror, and a juror’s age is positively correlated with the probability of the defense removing the juror. A black potential juror was also found much less likely to be removed if the defendant was also black. There was no statistical or practical significance found regarding the effect of a juror’s gender on peremptory challenges.
Literature review
Voir dire is the process by which defense and prosecution attorneys and the judge strike and replace potential jurors to select a jury. The judge, prosecuting attorney, and defense attorneys will ask potential jurors questions regarding their ability to be impartial. Some questions are routinely asked, such as ones to determine whether or not the potential juror is related to the defendant, or whether or not they feel they have a bias that would cause them to favor one side or the other (Neilson and Winter 2000). The answers to these and other questions will allow attorneys to remove potential jurors “for cause” (Neilson and Winter 2000). Removals for cause are unlimited and cannot be revoked by a trial judge, as they are directly protected under one’s Sixth Amendment right to an impartial jury (Beck 1998).
After for cause removals, attorneys may use a limited number of peremptory challenges to remove potential jurors. Peremptory challenges require no explanation and were historically unregulated, until in Batson v. Kentucky (476 U.S. 79 [1986]) and J.E.B. v. Alabama ex. rel. T.B. (511 U.S. 127 [1994]), the Supreme Court has ruled that using peremptory challenges for reasons based solely on race or gender is illegal. The number of peremptory challenges allowed to each side depends on the jurisdiction and on the severity of the crime. For minor crimes it may be that each side receives three to six challenges, while for capital crimes it is common that each side receives at least twenty challenges (Flanagan 2015). The purpose of these challenges is to allow attorneys to remove jurors that may be biased for reasons not obviously stated in the “for cause” removal step of voir dire.
Batson v. Kentucky and J.E.B. v. Alabama attempt to stop attorneys from using peremptory challenges to discriminate potential jurors based on race and gender, as these challenges would violate the equal protection clause of the U.S. Constitution. In Batson v. Kentucky, the defendant claimed the prosecutor violated his rights by using peremptory challenges to exclude all four black potential jurors during voir dire; these peremptory challenges led to an all-white jury that found the defendant guilty (Beck 1998). In response, the Supreme Court created a test to determine whether or not a peremptory challenge is used in an unconstitutional way, based either solely on race or gender. As Beck explains, the defendant must show that the circumstances of a peremptory challenge create “a prima facie case that the prosecutor challenged the potential juror on the basis of race” (1998). After this, the proponent of the peremptory challenge must provide a non-race based reason for exercising the challenge. The trial court then uses the information from this process to determine whether the opponent of the strike has proven purposeful discrimination (Beck 1998). From reading peremptory challenge guides from all types of sources online, it is clear that the use of juror demographics to predict conviction rates has been used as a strategy for attorneys to remove unfavorable potential jurors.
Zeisel et al. found that there is variation in the effectiveness in peremptory challenges, meaning some attorneys were better than others at using the challenges to their advantage (1978). It was also the first study that found peremptory challenges have had a significant effect on a case verdict. Since the Ziesel et al. study, predicting a juror’s likelihood to convict using demographic information has been studied further and many models have proven the importance of peremptory challenges on a case verdict. While the purpose of peremptory challenges lies in allowing attorneys to remove potential jurors with biases, peremptory challenges may be a tempting way to remove unbiased jurors that you predict will vote to acquit or convict.
Flanagan finds that the use of peremptory challenges may increase “the probability of juries composed entirely of members on one extreme or another of some ideological spectrum” (2015). Although peremptory challenges have been proven to eliminate jurors with “extreme biases,” the challenges are used strategically by each attorney. Therefore, the jurors kept will likely be correlated with each other through similar biases or characteristics (Flanagan 2015). The establishment of an impartial jury is important, as a jury that is biased is likely to make less effort to process information produced at trial and jump more quickly to judgment (Brunell et al. 2009).
Bognar et al. use a model to analyze juror behavior and the way biases can affect jury dynamics and ultimately the probability of conviction (2018). Bognar et al. argue that fewer peremptory challenges would limit attorneys’ ability to exclude highly opinionated jurors, and reduce bias. However, Bognar et al. find that greater polarization (when jurors are not too bias) actually leads to better informed verdicts (2018). As a result, they conclude that less peremptory challenges would lead to better jury decisions. This is very different from most previous literature, which state that removing bias jurors with peremptory challenges is beneficial to the decision-making process.
Flanagan (2018) finds that black male defendants are less likely to have a jury containing multiple black male jurors and therefore less likely to be tried by a “jury of their peers.” Flanagan finds that juries comprised of more black men are less likely to convict any defendant, especially black defendants. Flanagan also finds that each state peremptory challenge is associated with a 2.4-2.8% increase in the conviction rate for black men. This research is very significant because it shows that peremptory challenges can not only affect the makeup of a jury but conviction rates as well. This research contributes a large development on concepts Zeisel et al (1978) by empirically showing that state attorneys are using peremptory challenges to their advantage aggregate level.
Anwar et al. use models to show that black jurors are in fact more likely to favor the defense, and that one black potential juror, even if the jury does not have a black member, can have significant effects on a conviction rate (2010). Therefore, the presence of a black potential juror makes it more likely that the final jury will also favor the defense (Anwar et al. 2010). For this reason, even after Batson v. Kentucky, black jurors are disproportionately struck by the state.
On the other hand, the relevant Supreme Court cases do not limit discrimination based on any other characteristic including age. Anwar et al. (2014) find that the state is more likely to use a peremptory challenge to remove a younger juror and that older jurors are more likely to convict. Anwar et al. (2014) find that the discrimination used by attorneys is strategic and not based on false stereotypes. Likewise, they find that conviction rates increase by about 1 percentage point for each 1-year increase in the average age of the jury pool (2014). This is significant because discrimination based on age during peremptory challenges is not illegal, and it is proven that the results of these peremptory challenges may have major effects on conviction rates.
Schwartz and Schwartz (1996) find that the allowance of many peremptory challenges does help attorneys remove potential jurors with extreme biases, but sacrifices a representative jury. They also find that the two relevant Supreme Court cases limiting discrimination against potential jurors solely based on race or gender guarantee a certain amount of dissension among jurors, which increases the likelihood of a longer and more thoughtful deliberation (Schwartz and Schwartz 1996). They also argue that the rulings of the two Supreme Court cases are only relevant if relevant attitudes toward criminal conduct vary systematically between groups within the society, and the discrimination attorneys make use of are not based on false stereotypes. This is concerning because it could lead to a jury with underrepresented minority groups where the distinct points of view of certain minority groups are not represented. Schwartz and Schwartz discuss the idea of discriminatory peremptory challenges from both sides offsetting each other. The idea of discriminatory peremptory challenges refers to the concept that if one side discriminates against black potential jurors and the other side discriminates against white potential jurors, the peremptory challenges will cancel out and the jury will be left looking similar to the original jury pool, but with potential jurors removed that have particularly strong biases. The problem with this, as Schwartz and Schwartz explain, is that these challenges often do not cancel out and all-white juries (such as in Batson v. Kentucky) or all-women juries (such as in J.E.B. v. Alabama) may exist. In these cases, minority race jurors and men, respectively, were underrepresented in the jury pool itself, which meant discriminatory peremptory challenges had a very impactful effect on the final jury makeup (Schwartz and Schwartz 1996)
Data
The data used for this paper was collected by the North Carolina Jury Sunshine Project (NCJSP) at the Wake Forest University School of Law. The data is made up of information regarding felony trials across a large number of jurisdictions in North Carolina. All trials observed are non-capital, end in a verdict, have single defendants that are male and either black or white, and have complete information about how potential jurors were disposed. Two data sets were available to me, one containing aggregate data on the jury level, and another on the individual level, which has demographic information on potential jurors including race, age, and gender. I chose to use the individual level data, which contained 6,535 observations from 2011. However, since 161 observations were missing data on age, only 6,374 observations were used in the linear probability models. The data was collected using trial records and by matching the names of potential jurors with demographic information from the public voting registry.
Figure 1: Description of Variables
| Variable | Description |
| state_chal | Dummy variable. State used a peremptory challenge to excuse the juror = 1 |
| defense_chal | Dummy variable. Defense used a peremptory challenge to excuse the juror = 1 |
| age | Potential juror’s age in years |
| female | Dummy variable. If potential juror is female = 1 |
| black | Dummy variable. If potential juror is black = 1 |
| hispanic | Dummy variable. If potential juror is Hispanic = 1 |
| asian | Dummy variable. If potential juror is Asian = 1 |
| native | Dummy variable. If potential juror is Native American = 1 |
| other | Dummy variable. If potential juror’s race is indicated as “other” = 1 |
| def_black | Dummy variable. If defendant is black = 1 |
| blackdef_black | Interaction variable. If potential juror is black and defendant is black = 1 |
| def_white | Dummy variable. If defendant is white = 1 |
| blackdef_white | Interaction variable. If potential juror is black and defendant is white = 1 |
I chose to use age, female, black, hispanic, asian, native, and other as variables because race, age, and gender are mentioned in previous literature as observable characteristics that may be used as bias indicators for peremptory challenges. The race of the defendants is also taken into account through the variables def_white and def_black. The two outcome variables used in my linear probability models are state_chal and defense_chal, which indicate whether a potential juror was removed by a peremptory challenge from the state or defense. Finally, I included two interaction variables, blackdef_black and blackdef_white, to observe whether the race of the defendant has an effect on the magnitude of the effect of the juror’s race.
Data regarding potential jurors’ political affiliations were available to be but not included in the regression as they yielded no statistical significance. When interacted with other variables related to race and age, there was still no statistical significance and the addition of these variables did not affect the regression in a meaningful way. Additionally, the following interactions yielded no statistical significance and were omitted as a result: age*def_white, age*def_black, age*female, hispanic*def_black, asian*def_black, native*def_black, other*def_black, hispanic*def_white, asian*def_white, native*def_white, other*def_white. I believed these interaction variables may be interesting because the defendant’s race may affect the magnitude of the effect of other demographic variables, but the results were neither statistically significant not practically significant. Other coefficients in the regression were also unaffected by the addition of these variables.
Figure 2: Summary of Variables
| Variable | Obs | Mean | Std. Dev. | Min. | Max. |
| state_chal | 6,535 | 0.117368 | 0.3218829 | 0 | 1 |
| defense_chal | 6,535 | 0.179189 | 0.3835398 | 18 | 90 |
| age | 6,374 | 48.13618 | 13.74699 | 0 | 1 |
| female | 6,535 | 0.5562357 | 0.4968655 | 0 | 1 |
| black | 6,535 | 0.2179036 | 0.4128531 | 0 | 1 |
| hispanic | 6,535 | 0.0047437 | 0.0687161 | 0 | 1 |
| native | 6,535 | 0.0024484 | 0.049424 | 0 | 1 |
| asian | 6,535 | 0.00658 | 0.0808558 | 0 | 1 |
| other | 6,535 | 0.013619 | 0.1159118 | 0 | 1 |
| def_black | 6,535 | 0.6706963 | 0.4699964 | 0 | 1 |
| blackdef_black | 6,535 | 0.1640398 | 0.37034 | 0 | 1 |
| def_white | 6,535 | 0.3063504 | 0.4610123 | 0 | 1 |
| blackdef_white | 6,535 | 0.0463657 | 0.210292 | 0 | 1 |
Figure 2 show 11.73% of jurors were excluded through a peremptory challenge by the state, while 17.92% of the jurors were excluded through a peremptory challenge by the defense. This indicates that the defense uses more peremptory challenges than the state. This is interesting because in non-capital cases in North Carolina, each side is limited to six peremptory challenges per defendant. Since the data only includes information from non-capital trials with one defendant, I find that the defense was more likely to use a peremptory challenge to exclude a juror than the state.
The average juror age was about 48 years, 55.62% of jurors observed were female, 21.79% were black, 0.47% were Hispanic, 0.24% were Native American, 0.66% were Asian, and 0.014% of jurors had a race indicated as “other.”
Additionally, 67.07% of defendants were black and 30.64% of defendants were white. Most variables used in my model are dummy variables with a minimum value of 0 and a maximum of 1 with the exception of age, which has a minimum of 18 and a maximum of 90.
Methodology
I chose to use linear probability models to measure discrimination in peremptory challenges because it allows for the analysis of specific demographic characteristics of jurors; it also allows for the measure of a demographic’s likeliness to be used to determine whether or not an attorney will exclude a juror using a peremptory challenge.
This paper contains the results of four linear probability models. Two regressions use state peremptory challenges as the output variable, while the other two use defense challenges as the output variable. A juror removed by a peremptory challenge used by the state is considered a “success” when the output variable is state_chal, while a juror removed by a peremptory challenge used by the defense is considered a “success” when the output variable is defense_chal. It is important to use these output variables because state attorneys want to exclude potential jurors who are less likely to convict and defense attorneys want the opposite. Because there are well-established stereotypes regarding which demographic groups are more likely to convict, we can expect that most variables will have a positive coefficient for either the state or the defense, and a negative coefficient for the obverse. This is true of all statistically significant variables in the four regressions.
The use of many dummy variables and the limited number of explanatory variables could lead to incomplete results and cause the zero conditional mean assumption to be violated, as there are undoubtedly variables missing from the regression and are correlated with existing explanatory variables, such as race, age, and gender. This will be discussed further in the result section of the paper.
The four regressions in this paper allow me to assess how much the explanatory variables regarding race, gender, and age determine whether or not the state or defense uses a peremptory challenge to exclude a potential juror.
The first regression observed is:
state_chal = β1age + β2female + β3black + β4asian + β5hispanic + β6native + β7other +
β8def_black + β9blackdef_black + u
This regression will allow me to see the probability of the state using a peremptory
challenge to exclude a juror based on the juror’s age, gender, race, or the defendant’s race. The interaction variable, blackdef_black, will also allow me to see how the race of the defendant affects the magnitude of the effect of the juror’s race.
The second regression is:
defense_chal = β1age + β2female + β3black + β4asian + β5hispanic + β6native + β7other +
β8def_black + β9blackdef_black + u
This regression will allow me to see the probability of the defense using a peremptory
challenge to exclude a juror based on the juror’s age, gender, race, or the defendant’s race. The interaction variable, blackdef_black, will also allow me to see how the race of the defendant affects the magnitude of the effect of the juror’s race.
The third regression is:
state_chal = β1age + β2female + β3black + β4asian + β5hispanic + β6native + β7other +
β8def_white + β9blackdef_white + u
This regression will allow me to see the probability of the state using a peremptory
challenge to exclude a juror based on the juror’s age, gender, race, or the defendant’s race. The interaction variable, blackdef_white, will also allow me to see how the race of the defendant affects the magnitude of the effect of the juror’s race.
The fourth regression is:
defense_chal = β1age + β2female + β3black + β4asian + β5hispanic + β6native + β7other +
β8def_white + β9blackdef_white + u
This regression will allow me to see the probability of the defense using a peremptory
challenge to exclude a juror based on the juror’s age, gender, race, or the defendant’s race. The interaction variable, blackdef_white, will also allow me to see how the race of the defendant affects the magnitude of the effect of the juror’s race.
I’ve avoided issues regarding perfect or multiple collinearity because most variables I used were binary and I avoided collinearity by omitting variables for a white juror. Additionally, the variable for a white defendant is omitted when a variable for a black defendant is included and vice versa.
Omitted variable bias is undoubtedly an issue, as my only explanatory variables contain information about the age, race, and gender of potential jurors. There are definitely other factors that affect the probability of the state or defense using a peremptory challenge to exclude a juror and are correlated with race, age, and gender. Some omitted variables include a juror’s religion, income, or education. These variables are correlated with race, age, and gender in predictable ways but it’s difficult to tell in which direction they’re correlated with state or defense peremptory challenges. For example, income is definitely correlated with race but whether higher income potential jurors would be prefered by the state or defense is unclear and not thoroughly studied in previous literature. These limitations make it difficult to predict the direction of possible omitted variable bias. Therefore, it will be difficult to tell if race, age, or gender, rather than unobserved factors correlated with these characteristics, cause the defense or state to use peremptory challenges to exclude a potential juror.
Results
Figure 3: Regression 1 Results (state peremptory challenges)
| state_chal | Coef. | Std. Err. | t | P>t | [95% Conf. | Interval] |
| age** | -0.0007625 | 0.0003028 | -2.52 | 0.012 | -.0013561 | -.0001688 |
| female | -0.0109714 | 0.0080088 | -1.37 | 0.171 | -.0266713 | .0047285 |
| def_black | -0.0110537 | 0.0086885 | -1.27 | 0.203 | -.0280862 | .0059788 |
| blackdef_black** | 0.0655282 | 0.0256074 | 2.56 | 0.011 | .0153292 | .1157273 |
| black*** | 0.0693909 | 0.0214988 | 3.23 | 0.001 | .0272461 | .1115358 |
| asian | -0.0211526 | 0.0404449 | -0.52 | 0.601 | -.1004382 | .058133 |
| hispanic | 0.0264474 | 0.0598439 | 0.44 | 0.659 | -.0908668 | .1437617 |
| native | 0.0290269 | 0.0833247 | 0.35 | 0.728 | -.1343176 | .1923714 |
| other* | 0.0663391 | 0.038924 | 1.7 | 0.088 | -.009965 | .1426433 |
| _cons*** | 0.1400097 | 0.0172977 | 8.09 | 0 | .1061004 | .173919 |
R2 = 0.0260
*** denotes significance at 1% level
**denotes significance at 5% level
*denotes significance 10% level
The variables found statistically significant in the first regression are: age, black, blackdef_black, and other.
All else being equal, a one year increase in the potential juror’s age decreases the probability of the state using a peremptory challenge to remove the juror by 0.076%. The variable age is statistically significant at the 5% level, but not practically significant, as even a 10 year increase in the juror’s age would only increase the probability of the state removing the potential juror by 0.76%, which would most likely not have a major effect on the makeup of a jury.
A black potential juror is 6.9% more likely to be challenged by the state, compared to a white potential juror. The variable black is statistically significant at the 1% level and very practically significant; I believe anything that increases the probability of a potential juror being removed by more than 5% is significant because it indicates factors that could have a significant impact on the makeup of a jury, especially on the aggregate. Additionally, any indication that the state is excluding jurors based solely on race would point to a violation of the defendant’s Sixth Amendment right to an impartial jury of their peers.
Next, if there is a black potential juror and a black defendant, it is 12.39% more likely that the state will use a peremptory challenge to exclude the black potential juror, compared to if they were a white juror with a white defendant. The variable blackdef_black is statistically significant in the 5% level. This is practically significant because the interaction variable proves that having a black defendant increases the magnitude of discrimination against black potential jurors. Once again, the use of peremptory challenges based on the race of potential jurors is strictly unconstitutional. Additionally, disproportionately removing black potential jurors from the jury trial of a black defendant would actively work against allowing a black defendant to have a jury of their peers.
The final variable from regression 1 that is statistically significant is other. Although the variable other is statistically significant at the 10% level, it is not practically significant as only 0.01% of the potential jurors observed were identified as the race “other” and there is a lack of information about the variable.
Figure 4: Regression 2 Results (defense peremptory challenges)
| defense_chal | Coef. | Std. Err. | t | P>t | [95% Conf. | Interval] |
| age*** | 0.0020845 | 0.0003397 | 6.14 | 0 | 0.0014187 | 0.0027504 |
| female | -0.0062737 | 0.0096186 | -0.65 | 0.514 | -0.0251295 | 0.0125821 |
| def_black*** | 0.0433091 | 0.0118399 | 3.66 | 0 | 0.020099 | 0.0665192 |
| blackdef_black*** | -0.0928745 | 0.0216149 | -4.3 | 0 | -0.135247 | -0.050502 |
| black*** | -0.0784311 | 0.0191652 | -4.09 | 0 | -0.1160014 | -0.0408608 |
| asian | -0.0842336 | 0.05254 | -1.6 | 0.109 | -0.1872296 | 0.0187624 |
| hispanic | -0.0895499 | 0.0552825 | -1.62 | 0.105 | -0.1979223 | 0.0188225 |
| native | -0.0725609 | 0.0822663 | -0.88 | 0.378 | -0.2338305 | 0.0887087 |
| other** | -0.0883869 | 0.0356659 | -2.48 | 0.013 | -0.1583041 | -0.0184698 |
| _cons*** | 0.0887568 | 0.0191464 | 4.64 | 0 | 0.0512234 | 0.1262902 |
R2 = 0.0346
*** denotes significance at 1% level
**denotes significance at 5% level
*denotes significance 10% level
The variables found statistically significant in the second regression are: age, black, def_black, blackdef_black, and other.
All else being equal, a one year increase in the potential juror’s age increases the probability of the defense using a peremptory challenge to remove the juror by 0.21%. The variable age is statistically significant at the 1% level and practically significant, as a 30 year increase in the juror’s age would increase the probability of the defense removing the potential juror by 6.3%. If a 30 year age difference can affect the probability of a juror serving by 6.3%, age could have a significant effect on the makeup of a jury. In addition, previous literature has proven that older jurors are more likely to convict, meaning these removals by the defense may have an effect on the jury’s likeliness to convict (Anwar 2014).
A black potential juror is 7.8% less likely to be challenged by the defense, compared to a white potential juror. The variable black is statistically significant at the 1% level and very practically significant because it could have a large effect on the makeup of a jury. Additionally, previous literature has indicated that black jurors are less likely to convict, and this coefficient may be proof that defense attorneys are aware of this research or stereotypes that black jurors are less likely to convict (Anwar 2010).
Next, the coefficient on def_black means the probability of the defense using a peremptory challenge to remove a potential juror is 4.3% higher if the defendant is black, compared to if the defendant were white. This is statistically significant at the 1% level. On the other hand, this result is not practically significant in itself, as 4.3% is not very high and the analysis of this variable without the interaction requires some speculation regarding the significance of the coefficient. This result means a defense attorney is more likely to use a peremptory challenge if the defendant is black. One explanation could be that peremptory challenges are often used based on stereotypes and there are more strategic ways to use peremptory challenges when the defendant is black than when the defendant is white.
Finally, if there is a black potential juror and a black defendant, it is 12.8% less likely that the defense will use a peremptory challenge to exclude the black potential juror, compared to if they were a white juror with a white defendant. The variable blackdef_black is statistically significant in the 1% level. This is practically significant because the interaction variable shows that the magnitude of the effect of a juror’s race is higher if the defendant and potential juror are both black. In this case, it shows that defense attorneys are even more likely to see a black juror as a benefit if the defendant is black as well.
The final variable from regression 2 that is statistically significant is other. Although the variable other is statistically significant at the 5% level, it is not practically significant for reasons mentioned in the analysis of regression 1.
Figure 5: Regression 3 Results (state peremptory challenges)
| state_chal | Coef. | Std. Err. | t | P>t | [95% Conf. | Interval] |
| age** | -0.0007475 | 0.0003031 | -2.47 | 0.014 | -0.0013416 | -0.0001534 |
| female | -0.0109106 | 0.0080066 | -1.36 | 0.173 | -0.0266061 | 0.004785 |
| def_white | 0.0096463 | 0.0087865 | 1.1 | 0.272 | -0.0075783 | 0.0268709 |
| blackdef_white*** | -0.0718151 | 0.0263636 | -2.72 | 0.006 | -0.1234967 | -0.0201335 |
| black*** | 0.1341247 | 0.0136846 | 9.8 | 0 | 0.1072982 | 0.1609511 |
| asian | -0.0209355 | 0.0404617 | -0.52 | 0.605 | -0.100254 | 0.058383 |
| hispanic | 0.0270473 | 0.0599795 | 0.45 | 0.652 | -0.0905328 | 0.1446273 |
| native | 0.0315532 | 0.0829135 | 0.38 | 0.704 | -0.1309851 | 0.1940916 |
| other* | 0.0662059 | 0.0389463 | 1.7 | 0.089 | -0.0101419 | 0.1425537 |
| _cons*** | 0.1288663 | 0.0164069 | 7.85 | 0 | 0.0967033 | 0.1610294 |
R2 = 0.0262
*** denotes significance at 1% level
**denotes significance at 5% level
*denotes significance 10% level
The variables found statistically significant in the second regression are: age, black, blackdef_white, and other.
All else being equal, a one year increase in the potential juror’s age decreases the probability of the state using a peremptory challenge to remove the potential juror by 0.075%. The variable age is statistically significant at the 5% level but not practically significant, as even a 10 year increase in the juror’s age would only decrease the probability of the state removing the potential juror by 0.75%.
A black potential juror is 13.4% more likely to be challenged by the state, compared to a white potential juror. The variable black is statistically significant at the 1% level and very practically significant for reasons mentioned in the analysis of the variable black in regression 1.
Next, if there is a black potential juror and a white defendant, it is 7.3% more likely that the state will use a peremptory challenge to exclude the black potential juror, compared to if they were a white juror with a black defendant. The variable blackdef_white is statistically significant in the 1% level. This is practically significant because a 7.3% change in probability of excluding a potential juror may change the makeup of the jury. Additionally, the interaction variable shows that the magnitude of the effect of a juror’s race is affected by the race of the defendant. In this case, the interaction variable has a negative coefficient meaning it causes a downward shift of the intercept.
The final variable from regression 2 that is statistically significant is other. Although the variable other is statistically significant at the 10% level, it is not practically significant for reasons mentioned previously.
Figure 6: Regression 4 Results (defense peremptory challenges)
| defense_chal | Coef. | Std. Err. | t | P>t | [95% Conf. | Interval] |
| age*** | 0.0020781 | 0.00034 | 6.11 | 0 | 0.0014115 | 0.0027447 |
| female | -0.0065273 | 0.0096196 | -0.68 | 0.497 | -0.025385 | 0.0123304 |
| def_white*** | -0.0464896 | 0.0119805 | -3.88 | 0 | -0.0699754 | -0.0230039 |
| blackdef_white*** | 0.1031441 | 0.0231483 | 4.46 | 0 | 0.0577656 | 0.1485226 |
| black*** | -0.1714581 | 0.0100613 | -17.04 | 0 | -0.1911817 | -0.1517345 |
| asian | -0.0861456 | 0.0526918 | -1.63 | 0.102 | -0.1894393 | 0.017148 |
| hispanic* | -0.0920478 | 0.0546041 | -1.69 | 0.092 | -0.1990902 | 0.0149945 |
| native | -0.0829212 | 0.0812177 | -1.02 | 0.307 | -0.2421353 | 0.0762929 |
| other** | -0.0891602 | 0.0357266 | -2.5 | 0.013 | -0.1591964 | -0.0191241 |
| _cons*** | 0.1328008 | 0.0181383 | 7.32 | 0 | 0.0972437 | 0.1683579 |
R2 = 0.0350
*** denotes significance at 1% level
**denotes significance at 5% level
*denotes significance 10% level
The variables found statistically significant in the second regression are: age, black, def_white, blackdef_white, and other.
All else being equal, a one year increase in the potential juror’s age increases the probability of the defense using a peremptory challenge to remove the potential juror by 0.21%. The variable age is statistically significant at the 1% level and practically significant, as a 30 year increase in the juror’s age would increase the probability of the defense removing the potential juror by 6.3%. This is practically significant for reasons mentioned in the analysis of regression 1. Essentially, a 6.3% increase in the probability of removal of a juror due to a 30 year age difference could lead to a significantly different jury composition in the aggregate.
A black potential juror is 17.1% less likely to be challenged by the defense, compared to a white potential juror. The variable black is statistically significant at the 1% level and very practically significant for reasons mentioned in the analysis of the previous 3 regressions.
Next, if there is a black potential juror and a white defendant, it is 11.5% less likely that the defense will use a peremptory challenge to exclude the black potential juror, compared to if they were a white juror with a black defendant. This variable blackdef_white is statistically significant in the 1% level. This is practically significant because an 11.5% change in probability of excluding a potential juror may change the makeup of the jury. Additionally, the interaction variable shows that the magnitude of the effect of a juror’s race is affected by the race of the defendant. In this case, the interaction variable has a positive coefficient, meaning it causes an upward shift of the intercept.
A Hispanic potential juror is 9.2% less likely to be challenged by the defense, compared to a white potential juror. The variable hispanic is statistically significant at the 10% level and is only statistically significant in this regression. This is practically significant because a 9.2% affect on the probability that a defense attorney will exclude a potential juror may lead to a jury with a different makeup, and ultimately a different conviction result. Any indication that the state is excluding potential jurors based solely on race would indicate a violation of the defendant’s Sixth Amendment right to an impartial jury of their peers. This would include excluding a potential juror for being Hispanic.
The final variable from regression 2 which is statistically significant is other. Although the variable other is statistically significant at the 5% level, it is not practically significant for reasons mentioned previously.
A major limitation for this paper is lack of data. More information regarding the defendants would allow me to interact, for example, the defendant’s age with the potential juror’s age. It would be interesting to be able to observe whether or not attorneys are less likely to remove potential jurors because they share demographics other than race with the defendant. If possible, I would also get information regarding the questions asked in voir dire. It would be interesting to analyze whether or not certain demographics answered questions differently. This would allow me to study whether there are systematic differences, which vary by group, in the way certain demographics feel about crime. This study, however, would involve getting information from potential jurors that did not even make it to the peremptory challenge step of the voir dire process.
Finally, there is a lot of randomness and unpredictability when studying the behaviors of attorneys in these challenges. The R² value for all four regressions is relatively low, meaning a small percentage (less than 4%) of variation in the use of peremptory challenges is explained by the OLS regression line. In the four regressions, R² = 0.026, 0.034, 0.0262, and 0.0350, respectively. The two regressions with slightly higher R² values are the regressions where the output variables are state peremptory challenges. I believe including more information such as demographic information about the defendant would help me explain more of the variation.
Conclusion
The lack of enforcement of regulation created in response to Batson v. Kentucky and J.E.B. v. Alabama deem them ineffective. State attorneys continue to disproportionately remove black potential jurors, and defense attorneys continue to remove older jurors with peremptory challenges. The severity of discrimination in the judicial system can be seen in the Batson v. Kentucky case, where Batson was convicted of burglary after the state attorney struck all four black potential jurors and left an all-white jury. Batson received a 20 year sentence. Anything that could unfairly increase one’s likelihood to be convicted is unconstitutional and could have major long-term effects on defendants’ lives.
The research presented in this paper finds that whether or not a potential juror is black is the single most statistically and practically significant variable observed that affects the probability of a potential juror being either removed or kept by an attorney. More specifically, a black potential juror is 6.9% more likely to be challenged by the state, compared to a white potential juror. If the defendant it also black, the magnitude of this effect is greater and it is 12.39% more likely that the state will use a peremptory challenge to exclude the black potential juror, compared to if they were a white potential juror with a white defendant. Almost the exact opposite effect was found with the defense, with a black potential juror being 7.8% less likely to be challenged by the defense, compared to a white potential juror. With a black potential juror and a black defendant, it was 12.8% less likely that the defense will use a peremptory challenge to exclude the black potential juror, compared to if they were a white potential juror with a white defendant.
Like Flanagan (2018), I find that older jurors are also more likely to be removed by the defense and that the state is less likely to remove older jurors. Using the results found in Flanagan (2018) and Anwar (2014) it appears that attorney behavior concerning peremptory challenges is in line with empirical evidence regarding which juror demographics are more or less likely to convict. These challenges are used very strategically and have a real effect on a jury makeup and conviction rates, in agreement with Schwartz and Schwartz (1996).
Beck (1998) asked the question: if “for cause” challenges are protected under the Sixth Amendment and used to remove jurors that are biased or cannot be impartial, are peremptory challenges put into place to make a jury more impartial or tip the scales? I believe the results found in this research and previous literature should lead to a serious reconsideration of peremptory challenges, possibly leading to the elimination of the challenges as a whole. One common argument for peremptory challenges is that it is effective in removing jurors with extreme biases, but Bognar et al. (2018) find that greater polarization as a result of less peremptory challenges would actually lead to better jury decisions.
Because peremptory challenges are not protected under the constitution and have been around for so long, it would be interesting to see how a lack of peremptory challenges would change the judicial system. Additionally, research comparing the makeup of a jury after the “for cause” step of voir dire to the makeup after the peremptory challenge step would be interesting on an aggregate level, as we could study how different juries would look if peremptory challenges did not exist.
Works Cited
Anwar, Shamena, Patrick Bayer, and Randi Hjalmarsson. “The Impact of Jury Race in Criminal
Trials.” Quarterly Journal of Economics 127, no. 2 (2010): 1017-055. doi:10.3386/w16366.
Anwar, Shamena, Patrick Bayer, and Randi Hjalmarsson. “A Fair and Impartial Jury? The Role
of Age in Jury Selection and Trial Outcomes.” The Journal of Law & Economics 57, no. 4 (November 2014): 1001-030. doi:10.3386/w17887.
Anwar, Shamena, Patrick Bayer, and Randi Hjalmarsson. “Politics in the Courtroom: Political Ideology and Jury Decision Making.” NBER Working Paper No. 21145, May 2015. doi:10.3386/w21145.
Beck, Coburn R. “The Current State of the Peremptory Challenge.” William & Mary Law
Review 39, no. 3 (1998).
Bognar, Katalin, Moritz Meyer-Ter-Vehn, and Lones Smith. “A Conversational War of
Attrition.” Review of Economic Studies 85, no. 3 (December 2018): 1897-935.
doi:10.2139/ssrn.2334503.
Bringewatt, John P. “Snyder v. Louisiana: Continuing the Historical Trend Towards Increased Scrutiny of Peremptory Challenges.” Michigan Law Review 108, no. 7 (2010): 1283.
Brunell, Thomas L., Chetan Dave, and Nicholas C. Morgan. “Factors Affecting the Length of
Time a Jury Deliberates: Case Characteristics and Jury Composition.” Review of Law &
Economics 5, no. 1 (2009). doi:10.2202/1555-5879.1334.
Flanagan, Francis X. “Peremptory Challenges and Jury Selection.” The Journal of Law and
Economics 58, no. 2 (2015): 385-416. doi:10.1086/684040.
Flanagan, Francis X. “Race, Gender, and Juries: Evidence from North Carolina.” The Journal of
Law and Economics 61, no. 2 (2018): 189-214. doi:10.1086/698193.
Neilson, William S., and Harold Winter. “Bias and the Economics of Jury Selection.” International Review of Law and Economics 20, no. 2 (2000): 223-50. doi:10.1016/s0144-8188(00)00028-4.
Schwartz, Edward P., and Warren F. Schwartz. “The Challenge of Peremptory Challenges.”
Journal of Law, Economics, & Organization 12, no. 2 (October 1996): 325-60.
Zeisel, Hans, and Shari Seidman Diamond. “The Effect of Peremptory Challenges on Jury and
Verdict: An Experiment in a Federal District Court.” Stanford Law Review 30, no. 3 (1978): 491. doi:10.2307/1228114.