A Meta-Analysis on the Relationship Between Character Education and Student Achievement and Behavioral Outcomes

Abstract

An extensive meta-analysis, including 52 studies, was undertaken on the relationship between character education and student achievement and behavioral outcomes. Additional analyses were done to determine whether the effects of character education differed by student grade level, locale, race, and so on. The results indicated that character education is associated with higher levels of educational outcomes, no matter what type of standardized or nonstandardized measure was employed. Character education was also related to higher levels of expressions of love, integrity, compassion, and self-discipline. Overall, character education had somewhat greater effects for children in high school rather than those who were in elementary school. The effects of character education did not differ by the race of the children. The significance of these results is discussed.

Keywords

educational reform educational policy urban education school improvement school reform

For centuries, character education played a central role in the Western K-12 curriculum (Krisjansson, 2015; Ryan & Bohlin, 1999). As Thomas Lickona (1993), in his article, “The Return of Character Education,” notes, “Character education is as old as education itself. Down throughout history, education has had two great goals: to help people become smart and to help them become good” (p. 6). Moreover, most of the founders of the modern day education system including Plato, Cicero, and the early Christians believed that developing loving, compassionate, and self-disciplined individuals was actually more important than creating intellectually sophisticated people (Brooks, 2011; Krisjansson, 2015; McClellan, 1999). Thomas Lickona (1991) propounds what is probably the most recognized definition in contemporary society when he states, “Character education is the deliberate effort to develop the virtues that enable us to lead fulfilling lives and build a better world” (p. 228). In spite of the educational foundation mentioned above, many teachers have become reluctant to give character instruction in the classroom (Ryan & Bohlin, 1999).

Research indicates that, technically, many teachers would ideally like to incorporate character instruction into their classroom, but generally, do not do so for two reasons. First, they perceive that because of the emphasis that contemporary schooling places on high stakes testing, there is dearth of time available to teach character (Jeynes, 2006; Siegal, 2009). Second, most of the public school instructors have not been adequately trained to teach character and, therefore, teachers do not feel adequately prepared to engage in this practice (Brunn, 2014; Ryan & Bohlin, 1999; Siegal, 2009). Some teachers try to touch slightly on moral or character lessons by showing an antidrug DVD or inserting a brief statement on right or wrong, but neither of these even comes close to what, for centuries, has been what is commonly known as character education (Jones, Lewis, & Reffitt, 2013; Purpel, 2005).

In addition to these obstacles, as Lapsley and Narvaez (2006) point out, there are some teachers who think that character instruction does not belong in the classroom. Lapsley and Narvaez (2006) correctly note that character is often “controversial” (p. 248). David Carlin (2000) similarly states that, “Schools can be effective moral teachers when they represent communities that are morally homogeneous. The trouble is America is no longer a morally homogeneous community” (p. 147). These advocates believe that this kind of teaching should be left to one’s parents and that schools should be virtually neutral when it comes to teaching values in the class (Lapsley & Narvaez, 2006; Lapsley & Power, 2005; Purpel, 2005). Instead, they believe that schools should direct all their attention toward improving student achievement and critical thinking (Oliver, 1996; Paul & Elder, 2002; Purpel, 2005). The advocates of this approach believe that the job of teachers must be focused on graduating intelligent young people and that teaching some sense of ethics is fine, only if it is an outgrowth of the development of critical thinking (Paul & Elder, 2002).

There is no question that the number of educators who think that teaching about character and morality does not belong in the classroom has prevented many full-fledged character educator programs from being used in the schools (Lickona, 2004; Oliver, 1996; Paul & Elder, 2002). The unfortunate result of this is that some proponents of character instruction, such as Lapsley and Narvaez (2006), have greatly loosened the definition of character instruction so that they include efforts that consist of influencing a small aspect of behavior, for example, reducing drug use or bullying.

Clearly, those academics who attempt to do this mean well and are trying to convince others that there is great momentum in the character education movement (Lapsley & Narvaez, 2006; Stengel & Tom, 2006). For example, Stengel and Tom (2006) claim that after years of neglect, character education is making a “comeback” (p. 16). However, in reality, these attempts accentuate the fact that true character education programs in public schools are not commonly practiced (Lickona, 2004). Moreover, it is generally not wise for academics to identify an effort as a character education program as such, if the school that has launched that particular initiative does not view it that way.

In spite of the fact that some teachers find that the current school environment makes it difficult to teach virtue, several factors have caused a resurgence in the interest in character education including (a) the rise in school shootings and violence in U.S. public schools and in many nations across the globe; (b) the realization that some of the world’s greatest problems are a result of peoples’ inherent inability to get along, rather than lack of intelligence (Carlston, 2004; Jones et al., 2013); and (c) more of an emphasis on the fact that teaching, by definition, is a moral work (Sanger, Osguthorpe, & Fenstermacher, 2013). Although various academics are noting that teaching is a moral work, in almost all those cases, they define the moral component very broadly and it is quite distinct from “character education” as it is defined in this study. For example, Barbara Stengel (2013) believes in a moral component, but only in the context of her assertion, “The cultivation of moral and professional judgment seems unnecessary in my students’ educational worldview. Their own goodness and purity is taken for granted” (p. 46).

It is important to note that, historically speaking, virtually all the leading educational architects for 2,300 years believed that instruction in character and virtue was the most vital part of education (Jones et al., 2013; Marrou, 1956; Middleton, 2004; Moore, 2005). The reason why there was so much consensus on this issue is because individuals from Cicero to Martin Luther King believed that the most dangerous people on the face of the earth were those who were highly intelligent and schooled, but were not virtuous (Cicero, 2001; Middleton, 2004; Palmer, 2001). Educators throughout the centuries believed that it was of utmost importance to create a society that was loving, compassionate, civil, and self-disciplined (Cicero, 2001; Jones et al., 2013; Marrou, 1956; Middleton, 2004). Without this foundation, having intelligent people was of limited value (Krisjansson, 2015). Given this emphasis on virtue, American schools had character instruction as the centerpiece of the curriculum from the early 1600s until 1963 (Lickona, 2004; Middleton, 2004).

In the case of the United States, in a series of three U.S. Supreme Court decisions in 1962 and 1963, voluntary prayer and Bible reading were removed from public schools (Lickona, 2004; Sikorski, 1993). Although the U.S. Supreme Court did not specifically state that they were removing moral instruction from U.S. public schools, by removing the Judeo–Christian foundation of that character training, their decisions involved the de facto jettisoning of character instruction from American public school classrooms (Jeynes, 2002a; Lickona, 2004; Sikorski, 1993). Building on this initial impact is the fact that all it would take is one parent complaining by insisting that if a teacher taught about love or forgiveness that this was somehow Christianity being taught in the schools (Jeynes, 2012; Lickona, 2004; Sikorski, 1993). Even though one does not have to be a Christian to emphasize traits such as love and forgiveness, schools would quickly go into a retreat mode and remove all semblance of moral education in public schools (Arthur, 2014; Sikorski, 1993).

Recent Calls for the Reintroduction of Character Education

During the mid- to late-1990s, in particular, there was a new call for the reintroduction of a nonsectarian approach to character education in schools (Berkowitz & Bier, 2005; Jones et al., 2013; Lickona, 2004). Several of these calls came from political figures such as President Bill Clinton on the Democratic side and former Secretary of Education, Bill Bennett, on the Republican side (Bennett, 1996, 2011; Neuhaus, 2001). In reality, such declarations were hardly new, but previously, they had primarily come from parents, ministers, and those who combined spiritual leadership and calls to virtue such as Martin Luther King (Jeynes, 2007). By the mid-1990s, politicians were so alarmed by the surge in crime, out-of-wedlock births that were tearing apart hope for inner city youth, and a pervasive illegal drug problem that they understood that something had to be done to reintroduce a strong sense of virtue in America’s youth (Arthur, 2014; Brunn, 2014). President Clinton called for a renewed emphasis on self-discipline, school uniforms, and respect for religious liberty in the schools as keys to this end (Neuhaus, 2001). Bill Bennett asserted that there were certain values that virtually every person in the nation cherished, unless one was a criminal or sociopath, and these common virtues could be taught in the classroom (Bennett, 1996, 2011).

President Clinton and Secretary Bennett’s efforts had such dramatic effects that the implementation of school uniforms in public schools increased substantially, and many educators embraced Clinton’s guidelines for how to teach religion in the classrooms in a way that was objective, as had actually been prescribed in the U.S. Supreme Court’s decisions of 1962 and 1963 regarding Bible and prayer in the schools (Neuhaus, 2001). Concurrently, there were attempts by some who espoused character instruction to, with the absence of prior guidelines for moral education that existed before the U.S. Supreme Court’s decisions of 1962 and 1963, attempt to provide clarity regarding the aspects of character education. The Character Education Partnership (now known as Character.org) developed Eleven Principles of Character Education (Lapsley & Narvaez, 2006). One of the problems that has been evident in assessing the efficacy of character education programs is that, as Lapsley and Narvaez (2006) admit, a number of the programs they assess “might also be considered examples of one or more of the 11 Principles . . . ” (p. 269). However, here again, the loose definition of character education, with just one of the 11 principles needed to meet the definition is not at all acceptable for worthwhile quantitative analysis to take place on effects of character education.

Admittedly, there has been resistance to the idea of character education among a considerable number of educators (Arthur, 2014; Brunn, 2014). The most adamant opposition has come from two groups in particular: those who insist on increased standardized testing and those who emphasize multiculturalism (Gatto, 2001; Henningfeld, 2008). First, those who emphasize standardized tests argue that public schooling should be almost entirely an academic enterprise and there is no time for the teaching of right and wrong (Gatto, 2001; Henningfeld, 2008). According to this line of thinking, if the United States is to compete with East Asian schools, the nation needs more academic emphasis and more standardized testing (Gatto, 2001; Henningfeld, 2008). The response by those who advocate character instruction is twofold: (a) With all the moral problems that the United States has, we do not have time to avoid teaching character education and (b) many of the East Asian systems of education modeled their school systems after the American paradigm of the 1870-1945 period, and part of that rubric was moral education, which the East Asian schools have kept and the American public schools have largely jettisoned (Jeynes, 2007; Khan, 1997). Marvin Berkowitz (1998) avers, “Only when children are developed both intellectually and ethically are they truly educated” (p. 21).

Second, multiculturalists often ask the question, “Whose values shall we teach?” as a reason to teach no values at all. To this character instruction, proponents respond by saying, “Our values and human values.” That is, there are certain values that virtually every human being believes should be taught in the schools. Unless one is a criminal or a sociopath, these are values that virtually everyone embraces, for example, honesty, sincerity, responsibility, and respect. A Gallup poll reveals that between 91% and 99% of Americans believe that qualities such as honesty, the golden rule, courage, sincerity, patriotism, and responsibility should be taught in public schools and that, depending on the character trait, these percentages are often near 100% (Arthur, 2014; Brunn, 2014; Lahey, 2013). Nevertheless, the multiculturalist objections are valid in a sense, because character education should probably avoid controversies in which there is so much cultural variation that there is not much consensus.

The Need for a Meta-Analysis on Character Education

Given that the inclusion of a strong character instruction program is no longer the standard practice in public schools and many teachers question its salience (Lickona, 2004), it is vital that a meta-analysis be undertaken to determine its overall efficacy and to gain insight into whether the effectiveness of character education varies by age and the type of program initiated.

There are three particular challenges that face any attempt to assess the effects of character instruction, that make undertaking a meta-analysis on this topic particularly important. First, there are limited opportunities to test the effectiveness of character instruction programs. This is especially a result of the U.S. Supreme Court’s de facto removal of character education from public schools in 1962 and 1963. Even with the recent increased interest in reactivating character education programs, there are only a limited number of schools implementing character education programs in a substantial way (Arthur, 2014; Krisjansson, 2015). Second, virtually all studies of moral instruction examine its effects using only a limited number of outcome variables (Lickona, 2004). A meta-analysis allows for the broadest assessment of the relationship between character education and the academic and behavioral effects of character education. Third, virtually all studies examining character education use localized samples, which make generalizing the conclusion to the general student population difficult and, perhaps, impossible. Conducting a meta-analysis will draw from a wide array of studies from all across the country and all around the world.

Nevertheless, it should be noted that whatever results emerge from the meta-analysis, they will likely be very conservative compared with the actual effects of character education. This is because most studies examine the influence of moral instruction over a short period of time. Consequently, the effect sizes that emerge are likely going to be considerable underestimations of the degree of influence that restoring character would have for the full gamut of the 13-year period of K-12 schooling (Salkind & Rasmussen, 2007). Estimating the impact of restoring such instruction must be done in the context of this important caveat.

Method

Research Methods and Data Analysis Plan for the Meta-Analysis on the Character Education

Analytical approach

This meta-analysis examined the relationship between character education and student academic and behavioral outcomes in a school context. The procedures employed to conduct the meta-analysis are outlined in this section (“Analytical approach”) and the following sections: “Data Collection Method,” “Statistical Methods and Effect Size Statistics,” “Study Quality Rating,” “Character Education and Student Outcomes,” and “Defining of Variables.” Each study included in this meta-analysis met the following criteria:

It needed to examine character education and the specific independent variables in a way that could be conceptually and statistically distinguished from other primary variables under consideration. For example, if a researcher examined character education in conjunction with other independent variables that could not be statistically isolated from the other features, the study was not included in the analysis.

It needed to include a sufficient amount of statistical information to determine effect sizes. That is, a study needed to contain enough information so that test statistics, such as those resulting from a t test, ANOVA, and so forth, were either provided in the study or could be determined from the means and measures of variance listed in the study.

If the study used a control group, it had to qualify as a true control group and, therefore, be a fair and accurate means of comparison. Moreover, if the research utilized a control group at some times but not others, only the former comparisons were included in the meta-analysis.

The study could be a published or unpublished study.

Character education needed to be studied in the school context. Due, in part, to the fact that the vast majority of Western public schools do not have programs of character instruction, the overwhelming majority of the studies that have been done on this topic have been completed in other contexts. For example, a number of studies have been done on the effects of character training in service learning environments, while offering community service, while being raised in families that emphasize character, and in professional or higher education character programs. However, the purpose of this meta-analysis is to provide guidance for teachers specifically on character education within the school context. It should also be noted that socioemotional learning was only included if it involved broad enough outcome measures to be considered a measure of character and if it was assessed in a school context.

Due to the nature of the criteria listed above, qualitative studies were not included in the analysis. Although qualitative studies are definitely valuable, they are difficult to code for quantitative purposes and any attempt to do so might bias the results of the meta-analysis.

Data Collection Method (Coding and Rater Reliability)

To obtain the studies used in the meta-analysis, a search was undertaken to locate the relevant studies on character education. The first procedures to be used to locate these studies involved a computer search using 60 research databases (e.g., Psych Info., ERIC, Dissertation Abstracts International, Wilson Periodicals, Sociological Abstracts) to find studies examining character instruction and/or training (see Appendix). The search terms were character education, character instruction, moral education, values, values education, virtue, virtue education, self-discipline, and many other similar terms. Reference sections from journal articles on the character education were also examined to find additional research articles. The research team obtained a total of 90 studies that addressed the relationship under study, and found 52 studies that had a sufficient degree of quantitative data to include in this meta-analysis. Among the 52 studies that possessed a sufficient degree of quantitative data to include in this meta-analysis, the total number of subjects was approximately 225,779.

A number of different characteristics of each study were included for use in this study. These characteristics included (a) report characteristics, (b) sample characteristics, (c) intervention type, (d) the research design, (e) the grade level or age of the students, (f) the outcome and predictor variables, (g) the attrition rate, and (h) the estimate of the relationship between character instruction and student academic and behavioral outcomes.

Report characteristics: Each study entry began with the name of the author of the study. Then, the year of the study was recorded, followed by the type of research report. Research reports were defined either as a journal article, book, book chapter, dissertation, master’s thesis, government, school or private report, conference paper, or other type of report.

Sample characteristics included the number of students sampled, their locations, and how they were selected, for example, via random selection, stratified random selection, or via advertisement.

Intervention type: The researcher recorded the experimental or procedural manipulation used, if any, to determine the effects of character education.

Research design: The studies in this meta-analysis were categorized into three basic types of designs. First, the researcher noted the studies that employed some type of manipulations to assess the effects of character education. The second type of design included studies that took cross-sectional measures of character education without utilizing any type of manipulation. The third type of design involved the calculation of a correlational coefficient between character education and student academic and behavioral outcomes.

For studies that employed a manipulation to assess the effects of character education, the following were recorded: (a) the length, frequency, duration, and total number of training sessions; (b) the method of training (workshop, individual meetings, phone calls, videotape, email communication, newsletter); (c) the type of behavioral or achievement-related outcome measure (e.g., standardized achievement test, nonstandardized achievement test, or class grades); (d) the unit of analysis (individual student or classroom) at which the effect size was calculated; and (e) the magnitude of the relationship between character instruction and the measured outcomes.

For the cross-sectional studies and correlation studies, if they were available, also recorded (a) the socioeconomic status of participants in the sample and (b) the types of behavioral and academic measures that were used.

The grade level or age of the students was coded, including means and standard deviations when they were available.

The outcome and predictor variables from each study were coded to include the different ways that the character education was measured.

Attrition rate: When available, the attrition rate of each study was coded.

The estimate of the relationship between character education and student scholastic and behavioral variables: The process of the effect size estimation is described in the next section.

Statistical Methods and the Effect Size Statistics

Effect sizes were computed from data in such forms as t tests, F tests, p levels, frequencies, and r values via conversion formulas provided by Glass and his colleagues (Glass, McGaw, & Smith, 1981). When results were not significant, studies sometimes reported only a significance level. In the unusual case that the direction of these not-significant results was not available, the effect sizes were calculated to be zero.

For studies with manipulations, the standardized mean difference was used to estimate the effect of character education. The d index (Cohen, 1988) is a scale-free measure of the separation between two group means. Calculating the d index for any comparison involved dividing the difference between the two group means by either their average standard deviation or by the standard deviation of the control group. In the meta-analysis, the researchers subtracted the experimental group mean from the control group mean and divided the difference by their average standard deviation. As a supplement to these analyses, the Hedges’s “g” measure of effect size was used (Hedges & Vevea, 1998). Because it employed the pooled standard deviation in the denominator, it customarily provided a more conservative estimate of effect size. Hedges also provided a correction factor that helped to adjust for the impact of small samples.

For studies that involved cross-sectional measures of the relationship between character education and achievement, the following procedures were undertaken. For those studies that attempted to statistically equate students on other variables, the preferred measure of relationship strength was the standardized beta weight, b. These parameters were determined from the output of multiple regression analyses. If beta weights could not be obtained from study reports, the most similar measures of effect (e.g., unstandardized regression weights) were retrieved.

For studies that involved cross-sectional measures but included no attempt to statistically equate students on third variables, the results from the t tests, F tests, and correlation studies provided by the researchers in the study were used. Probability values were used as a basis for computation only if the researchers did not supply any information on the test statistics just mentioned.

Calculating average effect sizes

A weighting procedure was used to calculate average effect sizes across all the comparisons. First, each independent effect size was first multiplied by the inverse of its variance. The sum of these products was then divided by the sum of the inverses. Then, 95% confidence intervals (CIs) were calculated. As Hedges and Vevea (1998) recommend, all the analyses were conducted using fixed-error assumptions in one analysis and applied random-error assumptions in the other.

Tests of homogeneity were completed on the overall character education variables to gain a sense of the consistency of specific character education measures across studies.

Study Quality Rating

Two researchers coded the studies independently for quality, the presence of randomization, and whether the definitional criteria of the achievement gap are met. Study quality and the use of random samples will be graded on a 0 (lowest) to 3 (highest) scale. Quality was determined using the following:

Did it use randomization of assignment?

Did it avoid mono-method bias?

Did it avoid mono-operation bias?

Did it avoid selection bias?

Did it use a specific definition of character education?

I calculated interrater reliability by computing percentage of agreement on the definition of character education, issues of randomization, and quality of the study. A supplementary analysis was done to include only those studies with quality ratings of 3 and also 2 and 3.

Character Education and Student Outcomes

This meta-analysis examined the relationship between character instruction with school outcomes from the pre-kindergarten to college freshman levels. This meta-analysis first (Research Question 1) addressed whether there is a statistically significant relationship between character education and pre-kindergarten to college freshman student achievement and behavioral outcomes. A second question assessed whether the effects of character education differed by the age of the student (Research Question 2). The third analysis (Research Question 3) specifically focused on the relationship between character education and outcomes for students of color, as well as for students of low–socioeconomic status (low-SES). The final analysis addressed the effects of character education on specific measures of achievement and behavior (Research Question 4).

Definition of Variables

Independent variable

For the purposes of this study, character education will be defined as instruction designed to enhance love, integrity, self-discipline, and compassion in the lives of youth.

Dependent variables

Academic achievement was defined by such measures as grade point average (GPA), standardized test scores, and other measures. Regarding the achievement tests just mentioned, there were not only overall measures but also specific assessments in mathematics, reading, science, and social studies (history, civics, and geography). Additional academic indicators, referred to as “other measures,” included assessments of whether a child had been left back a grade.

Behavioral variables included measures of love, integrity, self-discipline, compassion, and a variety of other lifestyle measures.

Measures of socioeconomic status, race, and gender were also taken.

Results

The results indicated that there is a statistically significant relationship between character education instruction and overall student outcomes. Overall, the results of the meta-analysis indicated that there is a relationship between character education for kindergarten through college freshman youth as expressed in academic and behavioral outcomes combined. The results presented here used analyses based on random-error assumptions. The rationale for presenting these results rather than those using fixed-error assumptions is to utilize analyses that yielded more conservative effect sizes (Hedges & Vevea, 1998). As one would expect, the analyses based on fixed-error assumptions yielded somewhat larger effect sizes.

The results of this study indicate the overall U.S. character education variable yielded a statistically significant outcome of .31 (p < .01), 95% CI = [.10, .52], of a standard deviation. Table 1 indicates that the effect sizes ranged from a high of 2.57 to a low of .08. It is interesting to note that although there were a number of individual studies in the meta-analysis that did not yield statistically significant results, all of the overall relationships were in the positive direction. This is a very unusual result and will be more fully addressed in the “Discussion” section. The studies with the smallest samples produced the most extreme effect sizes on either end, consistent with the “funnel” pattern ideal in effect sizes (Greenhouse & Iyengar, 1994).

Table 1.

Studies Included in the Meta-Analysis Listed by Author, Year of Study, Type of Document, Sample Size, and a Variety of Other Characteristics.

Study and year	Sample size	Distinctions of study	Grade or age of students	Effect size without sophisticated controls	Effect size with sophisticated controls
D. W. Johnson, Johnson, Dudley, Mitchell, and Fredrickson (1997)	198	Midwestern suburban middle schools	Grades 6-9	2.57	—
DiBiase (2010)	45	Canadian sample	Grade 5	1.70	—
Soriano, Franco, and Sleeter (2011)	29	Spanish and Romanian sample	Grades 9-12	—	1.16
D. W. Johnson, Johnson, Dudley, and Magnuson (1995)	144	Midwestern suburban elementary schools	Grades 3-5	1.02	—
Grady (1999)	50	Examined both behavior and attitudes	High school	.99	—
Mucherah, Lapsley, Miels, & Horton, (2004)	67	A thorough 10-week study	Grades 4 and 5	.93	—
Flay and Allred (2003)	93 schools	Examined achievement and behavior	Grades K-12	.72	.62
Jeynes (2009a)	400	1-year program	Grades 7-12	.60	—
Flay, Allred, and Ordway (2001)	123 schools	1-year program	Grades K-6	.54	—
Snyder et al., (2010)	100	Examined achievement and behavior, 1-year-long program	Grades K-6	.53	—
Lewis (2012)	1, 170	About an equal number of males and females	Grades 3-8	.53	—
Stephens (2004)	337	Examined behavioral outcomes	Grades 10-11	.51	.27
Twemlow et al., (2001)	101	4-year program	Grades 3 and 5	.46	—
Skaggs and Bodenhorn (2006)	17.025	Program lasted 3 years	Grades 5-8	.45	.04
Jeynes (2009b)	140	Middle school and high school students	Grades 6-12	.43	—
Yurkewicz (2009)	149	Examined achievement and behavior	Grades 8-12	—	.41
Benninga, Berkowitz, Kuehn, and Smith (2003)	120 schools	Examined behavioral outcomes	Elementary school	.40	—
Jeynes (2002a)	20,706	Diverse large sample	Grade 12	.40	—
Hofmann-Towfigh (2007)	719	Examined behavioral outcomes	Elementary, middle, and high school	.39	.25
McClowry, Snow, and Tamis-LeMonda (2005)	148	Inner city students	Grades 1 and 2	.37	—
Muscott, Mann, and LeBrun (2008)	28	Examined achievement and behavior	Grades K-12	.37	—
Coleman, Hoffer, and Kilgore (1982)	25,000	Diverse large sample	Grades 9-12	.36	.20
Jeynes (2002b)	20,706	Diverse large sample	Grade 12	.36	.18
Elias, Gara, Schuyler, Branden-Muller, and Sayette (1991)	250	2-year program	Grades 9-11	.35	—
Raudenbush and Bryk (1986)	25,000	Examined HSB data set	Grades 9-12	.33	.27
Hoffer, Greeley, and Coleman (1985)	25,000	Examined HSB data set	Grades 9-12	.32	.18
Marshall, Caldwell, and Foster (2011)	19,317	2- to 5-year program	Grades K-12	.31	—
S. K. Johnson et al. (2014)	28	All male sample	Ages 17-19	.30	—
Williams, Yanchar, Jensen, and Lewis (2003)	204	Program used throughout the school	Grade 12	.30	—
Foa, Brugman, and Mancini (2012)	664	Italian sample	Grades 9 and 11	.28	.30
Gray and Watson (2002)	334	Examined GPA	College freshmen	.29	—
Prince, Ho, and Hanson (2010)	645	Diverse student group	Grades 2-5	.26	—
Parker, Nelson, and Burns (2010)	5,853	Examined behavior	Grades 1-5	.25	—
Zsolnai (2002)	438	Hungarian students	Grades 6 and 10	.24	—
Miller, Kraus, and Veltkamp (2005)	303	A large percentage of African American students	Grade 4	.23	—
Seider, Gilbert, and Norick (2013)	488	Examined GPA and behavior	Grades 6-8	—	.21
K. A. Johnson (1999)	23,000	African American sample	Grade 12	.20	—
Renfro, Huebner, and Ritchey (2003)	420	Almost all White student sample	Grade 9	.19	—
Corrigan, Grove, Vincent, Chapman, and Walls (2007)	490	Low-SES students	Elementary, middle, and high school	—	.17
Holtzapple et al. (2011)	8, 350	Examined behavioral outcomes	Grades 7-12	.17	—
Aber, Jones, Brown, Chaudry, and Samples (1998)	5,053	New York City schools	Grades 2-6	.17	—
Harrington, Giles, Hoyle, Feeney, and Yungbluth (2001)	1,655	Included 14 middle Schools	Grades 6-8	.17	.12
Botvin et al. (2003)	1,090	From 20 suburban schools	Grades 3-6	—	.17
Battistich, Solomon, and Watson (1989)	1,333	Suburban program	Grades K-4	.16	—
Battistich (2003)	521	Mostly White middle class	Grades 3-6	—	.15
Beets et al. (2009)	1,714	Hawaiian schools	Grades K-6	—	.14
Botvin, Griffin, Daz, and Ifill-William (2001)	3,621	New York City schools	Grades 7 and 8	.14	—
Sherblom, Marshall, and Sherblom (2006)	5,750	Looked at both reading and math achievement	Grades 3 and 4	.13	—
Munoz and Vanderhaar (2006)	1,039	Summer program	Grades 3 and 5	—	.13
van der velden, Brugman, Boom, and Koops (2010)	622	Foreign group of students	Grades 8	.10	—
Bavarian et al. (2013)	1, 170	Focused on students of color	Grades 3-8	—	.10
Luo et al. (2011)	11, 635	Example is from China	Grades 7 and 10	.08	.08

Note. SES = socioeconomic status.HSB = High School & Beyond Nationwide Dataset

Table 2 summarizes the studies by average year of the study, sample size, quality of study, and the quality of the definition of character education. The average year of the study was 2004.0. About 81% of the studies took place from 2000 and afterward. The average sample size was 4,341.9. Among the categories listed, the largest number of studies (21) had a sample size of 1,000+ and sample sizes of 100 to 499 (17). The average rating for the definition of character education in each of the studies was a good deal higher than the middle of the midpoint of the range of ratings allowable, 0 to 3. The mean quality of definition for character education for the studies was 2.69. The average quality of each study was 1.88 with most (about 71%) of the studies being rated either 3 or 2.

Table 2.

Means for Measures Assessing the Quality of Study, Whether a Random Sample Was Used, Year of Study, and Sample Size for the 40 Studies Included in the Meta-Analysis.

	M	SD or percentage distribution	Range
Year of study	2004.0	2010-2014 = 14 2000-2009 = 28 1990-1999 = 6 1970-1989 = 4	1970-2014
Sample size	4,341.9	1,000+ = 21 500-999 = 17 100-499 = 19 1-99 = 5	28-25,000
Quality of study	1.88	3 = 14 2 = 23 1 = 10 0 = 5	0-3
Quality of study’s definition of character education	2.69	3 = 39 2 = 12 1 = 1 0 = 0	0-3
Random sample	1.50	3 = 19 2 = 9 1 = 3 0 = 21	0-3

Tests of homogeneity for character education indicated that the measures were relatively homogeneous when sophisticated controls were used (χ² = 3.09, p = ns) and when sophisticated controls were not included (χ² = 2.92, p = ns).

Effect Sizes for Character Education Overall (Research Question 1)

Table 3 lists the effect sizes that emerged for character education as a whole, addressed under Research Question 1. The first outcomes examined included both academic and behavioral ones combined. Statistically significant effect sizes emerged for character instruction. The effect size for the U.S. overall character education variable was .31 (p < .01), 95% CI = [.10, .52], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .17, (p < .05), 95% CI = [.02, .32]. The effect size for the U.S. + Foreign overall character education variable was .29 (p < .01), 95% CI = [.08, .50], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .16, (p < .05), 95% CI = [.01, .31].

Table 3.

Effect Sizes for Character Education With 95% Confidence Intervals.

Type of overall character education variable	Effect size without sophisticated controls	Effect size with sophisticated controls	Overall effect size
Overall character education generally
U.S. overall	.31** [.10, .52]	.17* [.02, .32]	.25^a
U.S.+ foreign overall	.29** [.08, .50]	.16* [.01, .31]	.24^a
U.S. overall for studies rated 3	.33** [.11, .55]	.20* [.03, .37]	.28^a
U.S.+ foreign overall for studies rated 3	.33** [.11, .55]	.20* [.03, .37]	.28^a
U.S. overall for studies rated 2-3	.29** [.11, .55]	.17* [.02, .32]	.25^a
U.S.+ foreign overall for studies rated 2-3	.28** [.08, .50]	.16* [.01, .31]	.24^a
U.S. programs overall	.31** [.10, .52]	.18* [.02, .34]	.26^a
U.S.+ foreign programs overall	.30** [.09, .51]	.17* [.02, .32]	.25^a

Confidence intervals tabulation not undertaken for combined effect size because of difference in sample distributions for the two sets of studies.

p < .05. **p < .01. ***p < .001.

When the general overall measures were limited to studies with a quality rating of 3, the effect size for the U.S. overall character education variable was .33 (p < .01), 95% CI = [.11, .55], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .20 (p < .05), 95% CI = [.03, .37]. The effect size for the U.S. + foreign overall character education variable was .33 (p < .01), 95% CI = [.11, .55], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .20 (p < .05), 95% CI = [.03, .37].

When the general overall measures were limited to studies with a quality rating of 2 to 3, the effect size for the U.S. overall character education variable was .29 (p < .01), 95% CI = [.09, .49], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .17 (p < .05), 95% CI = [.02, .32]. The effect size for the U.S. + foreign overall character education variable was .28 (p < .01), 95% CI = [.07, .49], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .16 (p < .05), 95% CI = [.01, .31].

The effect size for U.S. overall character education programs was .31 (p < .01), 95% CI = [.10, .52], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .18 (p < .05), 95% CI = [.02, .34]. The effect size for the U.S. + foreign overall character education variable was .29 (p < .01), 95% CI = [.09, .49], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .17, (p < .05), 95% CI = [.02, .32]. The effects from programs were quite similar for the overall measures for character education, in part, because most of the character instruction that was examined was in the form of programs.

The meta-analytic outcomes for academic achievement specifically tended to be somewhat smaller than for the overall results for scholastic measures and behavior combined. The effect size for U.S. character education on achievement was .26 (p < .05), 95% CI = [.04, .48], of a standard deviation, which was statistically significant at the .05 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .17 (p < .05), 95% CI = [.02, .32]. The effect size for the U.S. + foreign overall character education variable was .24 (p < .05), 95% CI = [.03, .45], of a standard deviation, which was statistically significant at the .05 level of probability, when no sophisticated controls were used. The effect size was not statistically significant when sophisticated controls were used (.15, p < ns).

In contrast to the results just given, those for student behavior specifically tended to be somewhat larger than for the overall results for scholastic measures and behavior combined. The effect size for U.S. character education on achievement was .37 (p < .01), 95% CI = [.11, .63], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was also statistically significant when sophisticated controls were used, .18, (p < .05), 95% CI = [.02, .34]. The effect size for the U.S. + foreign overall character education variable was .36 (p < .05), 95% CI = [.10, .62], of a standard deviation, which was statistically significant at the .01 level of probability, when no sophisticated controls were used. The effect size was not statistically significant when sophisticated controls were used (.16, p < ns).

Effect Sizes for Character Education by Student Age (Research Question 2)

The effect sizes for character education varied considerably by age, especially in the academic achievement measures (see Tables 4 and 5). The results for high school students were larger for either elementary or middle school students. For the overall results of character education, they were .35 (p < .01), 95% CI = [.11, .63], of a standard deviation for U.S. students, and .33 (p < .01), 95% CI = [.11, .63], of a standard deviation for U.S. plus foreign students, when sophisticated controls are not used, both of which were statistically significant at the .01 level of probability. The results of character education for elementary school were smaller at .27 (p < .05), 95% CI = [.04, .50], of a standard deviation for U.S. students and .27 (p < .01), 95% CI = [.04, .50], of a standard deviation for U.S. plus foreign students, when sophisticated controls not used, both of which were statistically significant at the .05 level of probability. The effects for character education programs were slightly lower for middle school students than they were for high school students. The effects were .34 (p < .01), 95% CI = [.10, .57] and .32 (p < .01), 95% CI = [.08, .56] for U.S.- and U.S. plus foreign studies, respectively, for studies that did not use sophisticated controls.

Table 4.

Effect Sizes for Character Education With 95% Confidence Intervals.

Type of overall character education variable	Effect size without sophisticated controls	Effect size with sophisticated controls	Overall effect size
General effects for character education
U.S. academic achievement	.26* [.04, .48]	.17* [.02, .32]	.21^a
U.S. + foreign academic achievement	.24* [.02, .46]	.15	.19^a
U.S. student behavior	.37** [.11, .63]	.18* [.02, .34]	.30^a
U.S. + foreign student behavior	.36** [.10, .62]	.16	.29^a
U.S. elementary school students	.27* [.04, .50]	.12	.22^a.
U.S. + foreign elementary school students	.27* [.04, .50]	.12	.22^a
U.S. middle school students	.34** [.10, .57]	.20* [.04, .36]	.29^a
U.S. + foreign middle school students	.32** [.08, .56]	.12	.27^a
U.S. high school students	.35** [.11, .59]	.20* [.03, .37]	.30^a
U.S. + foreign middle high school students	.33** [.09, .57]	.18* [.02, .34]	.27^a

Confidence intervals tabulation not undertaken for combined effect size because of difference in sample distributions for the two sets of studies.

p < .05. **p < .01.

Table 5.

Effect Sizes for Character Education With 95% Confidence Intervals.

Character education and specific age levels of students	Effect size without sophisticated controls	Effect size with sophisticated controls	Overall effect size
Specific effects for character education
For academic achievement
U.S. elementary school students	.27* [.04, .50]	.15	.16^a
U.S. + foreign elementary school students	.27* [.04, .50]	.14	.15^a
U.S. middle school students	.14	.11	.13^a.
U.S. + foreign middle school students	.13	.11	.13^a
U.S. high school students	.32** [.08, .56]	.22* [.04, .40]	.29^a
U.S. + foreign middle high school students	.28* [.04, .52]	.20* [.02, .38]	.27^a
Specific populations
Minority students	.36** [.10, .62]	.27* [.04, .50]	.33^a
Low-SES students	.36** [.11, .61]	.27* [.06, .48]	.33^a

Note. SES = socioeconomic status.

Confidence intervals tabulation not undertaken for combined effect size because of difference in sample distributions for the two sets of studies.

p < .05. **p < .01.

The differences in the effects for character education by age were especially evident in the academic achievement measures. For high school students, the effects were .32 (p < .01), 95% CI = [.08, .56], of a standard deviation for U.S. students, and .28 (p < .05), 95% CI = [.04, .52], of a standard deviation for U.S. plus foreign students, when sophisticated controls are not used. For middle school students and their younger counterparts in elementary school, the results were not statistically significant, although they were in the expected positive direction at .16 and .14 of a standard deviation unit, respectively. When sophisticated controls were employed, the outcomes were .22 (p < .05), 95% CI = [.04, .40], for U.S. students, and .20 (p < .05), 95% CI = [.02, .38], for U.S. plus foreign students. The effects for the elementary (.15 and .14) and middle (.11 and .11) school students were in the expected direction, but were not statistically significant.

Effect Sizes for Character Education for Minority Students (Research Question 3)

The relationship between character education and academic and behavioral outcomes overall for minority students were statistically significant both when sophisticated controls were not utilized and also when they were. The effects were .36 (p < .01), 95% CI = [.10, .62], of a standard deviation unit when sophisticated controls were not used and .27 (p < .05), 95% CI = [.04, .50], of a standard deviation unit when sophisticated controls were utilized. The results for the meta-analysis for low-SES students had very similar numbers, although the 95% CIs were a little bit tighter for low-SES students than they were for minority pupils. The results were .36 (p < .01), 95% CI = [.11, .61], of a standard deviation unit when sophisticated controls were not used and .27 (p < .05), 95% CI = [.06, .48], of a standard deviation unit when sophisticated controls were utilized.

Effects of Character Education on Specific Measures of Achievement and Behavior (Research Question 4)

When one examines the specific aspects of academic achievement and student behavior, the effects of character education remain quite evident. The relationship between character education and scholastic outcomes yields statistically significant results whether one examines GPA, standardized tests, or more subjective measures, when no sophisticated controls are utilized. The effect sizes range from .25 to .41 for these three different academic measures. The effects were somewhat smaller when sophisticated controls were used versus when they were not for both standardized tests .21 (p < .01), 95% CI = [.02, .40], and nonstandardized measures, other than GPA, .26 (p < .01), 95% CI = [.02, .50]. Once sophisticated controls were employed, the results for the relationship between character education and GPA were still in the expected direction (.15), but were no longer statistically significant.

Table 6 indicates that the length of time character education was implemented also was related to larger effect sizes. Those efforts that were in place over a year in the United States yielded effects of .37 (p < .01), 95% CI = [.12, .62], of a standard deviation unit when sophisticated controls were not in place and .22 (p < .05), 95% CI = [.02, .42], of a standard deviation unit when sophisticated controls were used. For U.S. plus foreign character education, the results were .36 (p < .01), 95% CI = [.10, .62], of a standard deviation unit when sophisticated controls were not in place, and .22 (p < .05), 95% CI = [.02, .42], of a standard deviation unit when sophisticated controls were used. These results were somewhat larger than for the character education initiatives overall.

Table 6.

Effect Sizes for Character Education With 95% Confidence Intervals.

Character education and outcomes	Effect size without sophisticated controls	Effect size with sophisticated controls	Overall effect size
Specific outcomes
U.S. academic achievement	.26* [.04, .48]	.17* [.01, .33]	.21^a
U.S. + foreign academic achievement	.24* [.02, .46]	.15	.19^a
U.S. GPA	.41* [.11, .71]	.15	.33^a
U.S. + foreign GPA	.36* [.06, .66]	.16	.30^a
U.S. standardized tests	.25* [.05, .45]	.21* [.02, .40]	.23^a
U.S. + foreign standardized tests	.23* [.03, .43]	.17* [.01, .33]	.21^a
U.S. nonstandardized assessments	.33* [.06, .60]	.26* [.02, .50]	.31^a
U.S. + foreign nonstandardized assessments	.33* [.06, .60]	.26* [.02, .50]	.31^a
U.S. length of program	.37** [.12, .62]	.22* [.02, .42]	.32^a
U.S.+ foreign length of program	.36** [.10, .62]	.22* [.02, .42]	.31^a
Reading outcomes	.50** [.09, .91]	.40* [.06, .74]	.45^a
Math outcomes	.47** [.08, .86]	.37* [.05, .69]	.42^a
Science outcomes	.41* [.05, .77]	.22	.32^a
Social studies outcomes	.45* [.06, .84]	.26	.36^a

Confidence intervals tabulation not undertaken for combined effect size because of difference in sample distributions for the two sets of studies.

p < .05. **p < .01.

The results are quite consistent across tests of different subject matter. Whether one examines overall outcomes (tests and grading) in reading, math, science, or social studies, the meta-analysis yields effect sizes of nearly one half of a standard deviation unit when sophisticated controls are not utilized. When sophisticated controls were used, the effects for math (.37) and reading achievement (.40) declined somewhat. However, it should be noted that when these controls were included, the level of probability dropped from a .01 level of statistical significance to .05. In the cases of both science and social studies achievement, although the results were still a pretty good size (.22 and .26, respectively), they were no longer statistically significant. Once again, this combination can likely be attributed to the relatively small number of studies that examined these outcomes.

Given that the relationship between character instruction and behavioral outcomes was stronger than one finds for that type of education and academic outcomes, it likely comes as no surprise that all the effects for the individual behavioral variables were in the expected direction (see Table 7 and Figure 1). Moreover, all but one of these nine variables reached statistical significance. The results for a greater extent of self-control and lower rates of violence reached statistical significance both when sophisticated controls were and were not used. In the case of self-control, the numerical results were .58 (p < .01), 95% CI = [.17, .99], of a standard deviation unit when sophisticated controls were not in place, and .54 (p < .05), 95% CI = [.09, .99], of a standard deviation unit when sophisticated controls were used. When the violence findings were examined, the effects were .59 (p < .01), 95% CI = [.14, 1.04], of a standard deviation unit when sophisticated controls were not in place and .30 (p < .05), 95% CI = [.02, .58], of a standard deviation unit when sophisticated controls were used.

Table 7.

Effect Sizes for Character Education With 95% Confidence Intervals.

Type of overall character education variable	Overall effect size	Effect size with sophisticated controls	Overall effect size
Character education and behavioral results
Self-control	.58** [.17, .99]	.54* [.09, .99]	.57^a
Respect	.73** [.20, 1.26]	NA	.73
Love	.45* [.10, .80]	NA	.45
Honesty	.42* [.09, .75]	NA	.42
Suspensions	.53* [.09, .97]	NA	.53
Violence	.59* [.14, 1.04]	.30* [.02, .58]	.48^a
Moral judgment	.33* [.02, .64]	.28	.25
Bad behavior	.31* [.03, .59]	NA	.31
Social skills	.44* [.07, .81]	NA	.44

Note. NA = not available.

Confidence intervals tabulation not undertaken for combined effect size because of difference in sample distributions for the two sets of studies.

p < .05. **p < .01. ***p < .001.

The other behavioral outcomes included in the meta-analysis that yielded statistically significant results only included enough data to run analyses that did not use sophisticated controls. The results indicated that character instruction was associated with a smaller number of suspensions, .53 (p < .05), 95% CI = [.09, .97]; higher levels of respect, .73 (p < .01), 95% CI = [.20, 1.26]; higher levels of love, .45 (p < .05), 95% CI = [.10, .80]; better social skills, .44 (p < .05), 95% CI = [.07, .81]; a greater incidence of honesty, .42 (p < .05), 95% CI = [.09, .975]; and few expressions of bad behavior, .31 (p < .05), 95% CI = [.03, .59]. Only moral judgment yielded effect sizes that albeit were in the expected direction, .33 (p < .05), 95% CI = [.02, .64], and .28, but were only statistically significant when not utilizing sophisticated controls.

Discussion

Research Questions 1 and 2 probably yield the most notable results of the meta-analysis.

Effect Sizes for Character Education Overall (Research Question 1)

The results of the study indicate that there is a clear relationship between character education and student outcomes overall. The overall relationship appears to be about .3 to .4 of a standard deviation, which, in academic terms, would be about .4 of a GPA unit on a 4-point grading scale.

There is some indication that the relationship is probably somewhat larger between character instruction and student behavior than that of education and achievement, especially for younger children. These findings are consistent with what one would expect, given that character instruction is inherently more concerned with the hearts and behavior of youth than they are with scholastic outcomes (Jeynes, 2011; Matera, 2001). Nevertheless, there are many examples of scholars who argue that increased virtue in students will lead to higher levels of self-discipline, a sense of purpose, determination, perseverance, and wise priorities that all tend to lead to success (Khan, 1997; Rae & MacConville, 2015). It is noteworthy that the Educational Testing Service, that is, the College Board, concluded that one of the key factors contributing to the 17 consecutive years of decline in average SAT scores was the decline of the Judeo–Christian ethic (Jeynes, 2003; Wirtz, 1977).

The effects for character education were slightly more robust in the U.S. sample versus the U.S. plus foreign sample, but because the differences were small and were not statistically significant from one another, it is difficult to conclude whether these small differences may reflect anything substantial at all. However, the possibility in differences in effect sizes for foreign and U.S. samples may be a subject worthy of future study.

Effect Sizes for Character Education by Student Age (Research Question 2)

Probably, the most interesting set of results were those that emerged by age. The effects for character education were the largest for high school students, the second largest for middle school students, and the smallest for elementary school students. This trend was especially evident for academic achievement measures. These results are especially salient given that the overwhelming percentage of efforts to place character instruction in the schools is made at the elementary school and kindergarten levels. These results challenge that strategy.

There are two likely explanations for these results: (a) Character education may simply have a greater impact on adolescent students than it does on younger students and (b) the results for character education may be so strong at the high school level because it has a cumulative effect and what the studies are catching is this aggregate effect. Regarding the first point, these findings could simply reflect the fact that adolescents face a wider array of ostensibly moral decisions that include actions regarding premarital physical intimacy, illegal drugs, alcohol consumption, and other consequential choices in which character training can be highly worthwhile and remunerative. In terms of the second point, a number of studies included in this meta-analysis that examined character education in high school had character instruction in place for a number of years. Given that this meta-analysis also indicated that longer initiatives were associated with stronger results, the possibility that the larger effects for high school students reflects, in part, a cumulative phenomenon appears likely.

Both the above possible explanations for the findings certainly make sense. In fact, the extent to which each argument appears logical is such that further research should be undertaken to determine the extent to which each of these factors are at work.

Effect Sizes for Character Education for Minority Students (Research Question 3)

The results indicate that the effects of character education for minority students yield effect sizes that are at least as large as for the general population. These results are encouraging because, naturally, if one is to aver that character education efforts are to be initiated, one wants to see that it can help the broadest spectrum of people possible. The fact that moral instruction apparently benefits both children of color and those of lower socioeconomic status supports the notion that implementing character education efforts would have a high degree of academic and behavioral utility.

Effects of Character Education on Specific Measures of Achievement and Behavior (Research Question 4)

One of the most notable patterns in the data is the extent to which character education is related to higher academic outputs across virtually all the major subjects and is also associated with nearly all types of positive behavioral outcomes. From this meta-analysis, it would appear that the relationship between character education and these kinds of scholastic and behavioral results is pretty pervasive. The fact that the effects for character education hold across so many specific scholastic and behavioral outcomes would appear to provide impetus both for further research on this topic and the implementation of character education programs. The meta-analysis also likely provides insight into why educators for well above 2,000 years have placed such confidence in the efficacy of character education.

Limitations of Study

The primary limitation of this meta-analysis, or any meta-analysis, is that it is restricted to analyzing the existing body of literature. Therefore, even if the researcher conducting the quantitative integrations sees ways the studies included could have been improved, there is no way to implement those changes. A second limitation of a meta-analysis is that the social scientist is limited to addressing the same research questions addressed in the aggregated studies. For example, it would be advisable to have certain outcome measures from all the studies included, but one can only aggregate the existing results.

Concluding Thoughts

The results of this study quite strongly suggest that teachers and leaders need to revisit the potential value of character instruction. There is a certain irony to the fact that in terms of further research on this topic, a much wider implementation of character education is probably required. A greater utilization of character instruction will likely yield three benefits. First, to the extent that practicing more character instruction appears to yield both academic and behavioral benefits, schools and society will likely become stronger in a variety of ways. Second, once this is accomplished, further study of moral instruction will be facilitated. Third, contemporary schools will act more consistently with strategies that are known to have benefitted schools and society for more than 2,000 years. The results of this meta-analysis make it difficult to argue against teaching love, compassion, responsibility, honesty, and integrity in the schools.

Footnotes

Appendix

List of Search Engines Used in the Meta-Analysis.

Abstracts in social gerontology	Ejournals	NetLibrary
Academic Search Complete	EBSCO Ejournals	Newspapers
ACLS Humanities E-Book Project	EconLit	Oxford Journals Online
ACM Digital Library	Education Index Retrospective: 1929-1983	Oxford Reference Online
Alt-Press Watch	Education Line	Primary Search
American Indian Experience	ERIC	PsycARTICLES
Annual Reviews	Factiva	PsycINFO
Anthropology Plus	Family and Society Studies Worldwide	Public Administration Abstracts
AnthroSource	Handbook of Latin American Studies Online	Public Affairs Index
AP Images (formerly AccuNet)	Historical Abstracts	Rand California
Association Unlimited	Latino Literature	SAGE Premier Journals Online
ATLA Religion Database	Lexis Nexis Academic	Science Citation Index (SCI) see Web of Science
Black Studies Center	Library Literature & Information Science (H. W. Wilson)	Social Science Citation Index (SSCI), see Web of Science
Brill’s New Jacoby Online	Library, Information Science & Technology Abstracts	Social Services Abstracts
Business Monitor Online	MAS Ultra—School Edition	SocINDEX
Chicano Database	MEDLINE (via OVID)	Sociological Abstracts
CINAHL Plus with Full Text	Military & Government Collection	SpringerLink Journals Online Collection
Communication & Mass Media Complete	Natural Standard Professional Database	SPORTDiscus
Dissertation Abstracts International	NetLibrary	Wiley InterScience (including Blackwell Synergy journals)
Dissertation & Theses	Newspapers	WorldCat

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biography

William H. Jeynes is a Senior Fellow at Princeton’s Witherspoon Institute and a Professor of Education at California State University, Long Beach. He specializes in conducting meta-analyses. He has given speeches for the White House and several U.S. Government departments. He graduated first in his class from Harvard University and has advised a number of U.S. and foreign leaders.

References

*Aber

J. L.

Jones

S. M.

Brown

J. L.

Chaudry

Samples

(1998). Resolving conflict creativity: Evaluating the developmental effects of a school-based violence prevention program in neighborhood and classroom context. Development and Psychopathology, 10, 187-213.

Arthur

(2014). Traditional approaches to character education in Britain and America. In Nucci

L. P.

Krettenauer

Narvaez

(Eds.), Handbook of moral and character education (pp. 43-60). New York, NY: Routledge.

*Battistich

(2003). Effects of a school-based program to enhance prosocial development on children’s peer relations and social adjustment. Journal of Research in Character Education, 1, 1-17.

*Battistich

Solomon

Watson

(1989). Effects of an elementary school program to enhance prosocial behavior on children’s cognitive-social problem-solving skills and strategies. Journal of Applied Developmental Psychology, 10, 147-169.

*Bavarian

Lewis

K. M.

DuBois

D. L.

Acock

Vuchinich

Silverthorn

. . . Flay

B. R.

(2013). Using social-emotional and character development to improve academic outcomes: A matched-pair, cluster-randomized controlled trial in low-income, urban schools. Journal of School Health, 83, 771-774.

*Beets

M. W.

Flay

B. R.

Vuchinich

Snyder

F. J.

Acock

(2009). Using a social and character development program to prevent substance abuse, violent behaviors, and sexual activity among elementary school students in Hawaii. American Journal of Public Health, 99, 1438-1445.

Bennett

W. J.

(1996). The book of virtues. New York, NY: Simon & Schuster.

Bennett

W. J.

(2011). The book of man: Readings on the path to manhood. Nashville, TN: Thomas Nelson.

*Benninga

J. S.

Berkowitz

M. W.

Kuehn

Smith

(2003). The relationship of character education implementation and academic achievement in elementary schools. Journal of Research in Character Education, 1, 19-32.

10.

Berkowitz

M. W.

(1998). A primer for evaluating: A character education initiative. Washington, DC: Character Education Partnership.

11.

Berkowitz

M. W.

Bier

(2005). The interpersonal roots of character education. In Lapsley

D. K.

Power

F. C.

(Eds.), Character psychology and character education (pp. 268-285). Notre Dame, IN: Notre Dame University Press.

12.

*Botvin

G. J.

Griffin

K. W.

Daz

Ifill-William

(2001). Drug abuse prevention among minority adolescents: Postest and one-year follow-up of a school-based prevention intervention. Prevention Science, 2, 1-13.

13.

*Botvin

G. J.

Griffin

K. W.

Paul

Macaulay

A. P.

(2003). Preventing tobacco and alcohol use among elementary school students through life skills training. Journal of Child & Adolescent Substance Abuse, 12(4), 1-18.

14.

Brooks

(2011). The social animal: The hidden sources of love, character, and achievement. New York, NY: Random House.

15.

Brunn

(2014). Pedagogy for the whole child. In Nucci

L. P.

Krettenauer

Narvaez

(Eds.), Handbook of moral and character education (pp. 263-271). New York, NY: Routledge.

16.

Carlin

D. R.

(2000). Schools cannot teach moral values. In Willims

Mary E.

(Ed.), Education: Opposing viewpoints (pp. 144-147). San Diego, CA: Greenhaven Press.

17.

Carlston

(2004). Surviving Columbine: How faith helps us find peace when tragedy strikes. Salt Lake City, UK: Deseret.

18.

Cicero

M. T.

(2001). On moral ends. Cambridge, UK: Cambridge University Press.

19.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.

20.

*Coleman

Hoffer

Kilgore

(1982). High school achievement: Public, catholic, and private schools compared. New York, NY: Basic Books.

21.

*Corrigan

M. W.

Grove

Vincent

P. F.

Chapman

Walls

R. T.

(2007). The importance of multidimensional baseline measurements to assessment of integrated character education models. Journal of Research in Character Education, 5, 103-129.

22.

*DiBiase

(2010). The impact of a psych-educational prevention program for behaviorally at-risk students, equip for educators. Journal of Character Education, 8, 21-59.

23.

*Elias

M. J.

Gara

M. A.

Schuyler

T. F.

Branden-Muller

L. R.

Sayette

M. A.

(1991). The promotion of social competence: Longitudinal study of a preventive school-based program. American Journal of Orthopsychiatry, 61, 409-417.

24.

*Flay

B. R.

Allred

C. G.

(2003). Long-term effects of the Positive Action Program. American Journal of Health Behavior, 27(Suppl. 1), S6-S21.

25.

*Flay

B. R.

Allred

C. G.

Ordway

(2001). Effects of the Positive Action Program on achievement and discipline: Two matched-control comparisons. Prevention Science, 2, 71-89.

26.

*Foa

Brugman

Mancini

(2012). School moral atmosphere and normative orientation to explain aggressive and transgressive behaviors at secondary school. Journal of Moral Education, 41, 1-22.

27.

Gatto

J. T.

(2001). The underground history of American schooling. New York, NY: Oxford Village Press.

28.

Glass

G. V.

McGaw

Smith

M. L.

(1981). Meta-analysis in social research. Beverly Hills, CA: SAGE.

29.

*Grady

F. A.

(1999). Cluster school: A study of the impact in adulthood of a moral education intervention project. Dissertation Abstracts International, Section A: Humanities and Social Sciences, 55(7A), 1808.

30.

*Gray

E. K.

Watson

(2002). General and specific traits of personality, and their relation to sleep and academic performance. Journal of Personality, 70, 177-206.

31.

Greenhouse

J. B.

Iyengar

(1994). Sensitivity analysis and diagnosis. In Cooper

Iyengar

(Eds.), Handbook of research synthesis (pp. 383-398). New York, NY: Russell Sage Foundation.

32.

*Harrington

N. G.

Giles

Hoyle

R. H.

Feeney

G. J.

Yungbluth

S. C.

(2001). Evaluation of the all stars character education and Problem Behavior Prevention Program: Effects on mediator and outcome variable for middle school students. Health Education & Behavior, 29, 533-546.

33.

Hedges

L. V.

Vevea

J. L.

(1998). Fixed and random effects models in meta-analysis. Psychological Method, 3, 486-504.

34.

Henningfeld

D. A.

(2008). Standardized testing. Detroit, MI: Greenhaven Press.

35.

*Hoffer

Greeley

A. M.

Coleman

J. S.

(1985). Achievement growth in public and Catholic schools. Sociology of Education, 58, 74-97.

36.

*Hofmann-Towfigh

(2007). Do students’ values change in different types of schools? Journal of Moral Education, 36, 453-473.

37.

*Holtzapple

C. K.

Griswold

J. S.

Cirillo

Rosebrock

Nouza

Berry

(2011). Implementation of a school-wide adolescent character education and prevention program: Evaluating the relationships between principal support, faculty implementation, and student outcomes. Journal of Character Education, 9, 71-90.

38.

Jeynes

(2002a). A meta-analysis of the effects of attending religious schools and religiosity on Black and Hispanic academic achievement. Education and Urban Society, 35(1), 27-49.

39.

Jeynes

(2002b). Educational policy and the effects of attending a religious school on the academic achieveement of children. Educational Policy,16(3), 406-424.

40.

Jeynes

(2003). A meta-analysis: The effects of parental involvement on minority children’s academic achievement. Education and Urban Society, 35(2), 202-218.

41.

Jeynes

(2006). Standardized tests and Froebel’s original kindergarten model. Teachers College Record, 108(1), 1937-1959.

42.

Jeynes

(2007). American educational history: School, society, and the common good. Thousand Oaks, CA: SAGE.

43.

Jeynes

(2009a). The relationship between Bible literacy and academic achievement and school behavior. Education and Urban Society, 41(4), 419-436.

44.

Jeynes

(2009b). The relationship between biblical literacy, academic achievement, and school behavior among Christian- and public- school students. Journal of Research on Christian Education, 18(1), 36-55.

45.

Jeynes

(2011). Help families by fostering parental involvement. Phi Delta Kappan, 93(3), 38-39.

46.

Jeynes

(2012). A meta-analysis on the effects and contributions of public, public charter, and religious schools on student outcomes. Peabody Journal of Education, 87(3), 305-335.

47.

*Johnson

D. W.

Johnson

R. T.

Dudley

Magnuson

(1995). The impact of conflict resolution training on middle school students. Journal of Social Psychology, 135, 673-686.

48.

*Johnson

D. W.

Johnson

R. T.

Dudley

Mitchell

Fredrickson

(1997). The impact of conflict resolution training on middle school students. Journal of Social Psychology, 137, 11-21.

49.

*Johnson

K. A.

(1999). Comparing math scores of black students in D.C.’s public and Catholic schools. Washington, DC: Heritage Foundation.

50.

*Johnson

S. K.

Hershberg

R. M.

Arbeit

M. R.

DeSouza

L. M.

Callina

K. S.

Gutierrez

A. S.

. . . Rubin

R. O.

(2014). Exploring characteristics of young adult men: Initial findings from mixed methods evaluation of an all-male, character focused trade school. Journal of Character Education, 10, 129-154.

51.

Jones

M. L.

Lewis

P. A.

Reffitt

K. E.

(2013). Toward human flourishing: Character, practical wisdom, and professional formation. Macon, GA: Mercer University Press.

52.

Khan

(1997). Japanese moral education: Past and present. Madison, WI: Farleigh Dickinson University Press.

53.

Krisjansson

(2015). Aristolelian character education. New York, NY: Abington.

54.

Lahey

(2013, May 6). The benefits of character education: What I learned from teaching at a “core virtues” school. The Atlantic. Retrieved from http://www.theatlantic.com/national/archive/2013/05/the-benefits-of-character-education/275585/

55.

Lapsley

D. K.

Narvaez

(2006). Character education. In Renninger

K. A.

Sigel

I. E.

(Eds.), Handbook of child psychology, Vol. 4: Child psychology in practice (pp. 248-296). Hoboken, NJ: John Wiley.

56.

Lapsley

D. K.

Power

F. C.

(2005). Moral psychology and the crossroads. In Lapsley

D. K.

Power

F. C.

(Eds.), Character psychology & character education (pp. 18-35). Notre Dame, IN: Notre Dame University Press.

57.

*Lewis

K. M.

(2012). Evaluation of a socio-emotional and character development program: Methods and outcomes. Dissertation Abstracts, 74(3-B), E.

58.

Lickona

(1991). Educating for character. New York, NY: Bantam Books.

59.

Lickona

(1993). The return of character education. Educational Leadership, 51(3), 6-11.

60.

Lickona

(2004). Character matters. New York, NY: Simon & Schuster.

61.

*Luo

Shi

Zhang

Liu

Rozelle

Sharbono

(2011). Community service, educational performance and social responsibility in Northwest China. Journal of Moral Education, 40, 181-202.

62.

Marrou

H. I.

(1956). A history of education in antiquity. New York, NY: Sheed & Ward.

63.

*Marshall

J. C.

Caldwell

S. D.

Foster

(2011). Moral education the CHARACTERplus way. Journal of Moral Education, 40, 51-72.

64.

Matera

(2001). A cry for character. Paramus, NJ: Prentice Hall.

65.

McClellan

B. E.

(1999). Moral education in America. New York, NY: Teachers College Press.

66.

*McClowry

Snow

D. L.

Tamis-LeMonda

C. S.

(2005). An evaluation of the effects of “insights” on the behavior of inner city primary school children. Journal of Primary Prevention, 26, 567-584.

67.

Middleton

(2004). Religion and education. In Ryan

M. C.

(Ed.), Living in colonial America (pp. 65-74). San Diego, CA: Greenhaven Press.

68.

*Miller

T. W.

Kraus

R. F.

Veltkamp

L. J.

(2005). Character education as a prevention strategy in school-related violence. Journal of Primary Prevention, 26, 455-466.

69.

Moore

J. P.

(2005). One nation under God: The history of prayer in America. New York, NY: Doubleday.

70.

*Mucherah

Lapsley

D. K.

Miels

Horton

(2004). An intervention to improve sociomoral climate in elementary school classrooms: An evaluation of Don’t Laugh at Me. Journal of Research in Character Education, 2, 45-58.

71.

*Munoz

M. A.

Vanderharr

J. E.

(2006). Literacy-embedded character education in a large urban district: Effects of the child development project on elementary school students and teachers. Journal of Research in Character Education, 4, 47-64.

72.

*Muscott

H. S.

Mann

E. L.

LeBrun

M. R.

(2008). Positive behavioral interventions and supports in New Hampshire: Effects of large-scale implementation of schoolwide positive behavior support on student discipline and academic achievement. Journal of Positive Behavior Interventions, 10, 190-205.

73.

Neuhaus

R. J.

(2001). The best of the “public square”: Book 2. Grand Rapids, MI: Eerdmans.

74.

Oliver

T. S.

(1996, March 25). Neutral and free: Religion, values, and the public schools. Sojourners, pp. 12-13.

75.

Palmer

J. A.

(2001). Fifty major thinkers on education. London: Routledge.

76.

*Parker

D. C.

Nelson

J. S.

Burns

M. K.

(2010). Comparison of correlates of classroom behavior problems in schools with and without a school-wide character education program. Psychology in the Schools, 47, 817-827.

77.

Paul

Elder

(2002). Critical thinking: Tools for taking charge of your professional and personal life. Upper Saddle River, NJ: Financial Times.

78.

*Prince

C. K.

E. A.

Hanson

S. B.

(2010). Effects of a school based program to improve adaptive school behavior and social competencies among elementary school youth: The Living Skills Program. Journal of Character Education, 8(2), 39-59.

79.

Purpel

D. E.

(2005). Moral outrage in education. New York, NY: Peter Lang.

80.

Rae

MacConville

(2015). Using positive psychology to enhance student achievement: A school-based programme for character education. Abingdon, UK: Routledge.

81.

*Raudenbush

Bryk

A. S.

(1986). A hierarchical model for studying school effects. Sociology of Education, 59, 1-17.

82.

*Renfro

Huebner

Ritchey

(2003). The effects of a university and high school partnership. Journal of School Violence, 2, 81-99.

83.

Ryan

Bohlin

K. E.

(1999). Building character in schools: Practical ways to bring moral instruction to life. San Francisco, CA: Jossey-Bass.

84.

Salkind

N. J.

Rasmussen

(2007). Encyclopedia of measurement and statistics. Thousand Oaks, CA: SAGE.

85.

Sanger

M. N.

Osguthorpe

R. D.

Fenstermacher

G. D.

(2013). The moral work of teaching in teacher education. In Sanger

M. N.

Osguthorpe

R. D.

(Eds.), The moral work of teaching and teacher education (pp. 3-13). New York, NY: Teachers College Press.

86.

*Seider

Gilbert

J. K.

Norick

(2013). The role of moral and performance character strengths in predicting achievement and conduct among urban middle school students. Teachers College Record, 115(8), 1-34.

87.

*Sherblom

S. A.

Marshall

J. C.

Sherblom

J. C.

(2006). The relationship between school climate and math and reading achievement. Journal of Research in Character Education, 4, 19-31.

88.

Siegal

(2009). The Oxford handbook of the philosophy of education. New York, NY: Oxford University Press.

89.

Sikorski

(1993). Controversies in constitutional law. New York, NY: Garland.

90.

*Skaggs

Bodenhorn

(2006). Relationships between implementing character education, student behavior, and student achievement. Journal of Advanced Academics, 18, 82-114.

91.

*Snyder

Flay

Vuchinich

Acock

Washburn

Beets

K.-K.

(2010). Impact of a social-emotional and character development program on school-level indicators of academic achievement, absenteeism, and disciplinary outcomes: A matched-pair, cluster randomized, controlled trial. Journal of Research of Educational Effectiveness, 3, 28-55.

92.

*Soriano

Franco

Sleeter

(2011). The impact of a values education program for adolescent Romanies in Spain on their feelings of self-realization. Journal of Moral Education, 40, 217-235.

93.

Stengel

B. S.

(2013). Teaching moral responsibility: Practical reasoning in a pedagogical “wonderland.” In Sanger

M. N.

Osguthorpe

R. D.

(Eds.), The moral work of teaching and teacher education (pp. 44-59). New York, NY: Teachers College Press.

94.

Stengel

B. S.

Tom

A. R.

(2006). Moral matters: Five ways to develop the moral life of Schools. New York, NY: Teachers College Press.

95.

*Stephens

J. M.

(2004). Motivation, morality, and academic (mis)conduct among adolescents. Stanford, CA: Stanford University.

96.

*Twemlow

S. W.

Fonagy

Sacco

F. C.

Gies

M. L.

Evans

Ewbank

(2001). Creating a peaceful school learning environment: A controlled study of an elementary school intervention to reduce violence. American Journal of Psychiatry, 158, 808-810.

97.

*van der Velden

Brugman

Boom

Koops

. (2010). Effects of EQUIP for educators on students’ self-serving cognitive distortions, moral judgement, and antisocial behavior. Journal of Character Education, 8, 77-95.

98.

*Williams

D. D.

Yanchar

S. C.

Jensen

L. C.

Lewis

(2003). Character education in a public high school: A multi-year inquiry into unified studies. Journal of Moral Education, 32, 3-33.

99.

Wirtz

(1977). On further examination. New York, NY: College Entrance Examination Board.

100.

*Yurkewicz

(2009). Happiness in adolescence: Contributions of self-control, hope, and personal strivings. New York, NY: Saint John’s University.

101.

*Zsolnai

(2002). Relationship between children’s social competence, learning, and school achievement. Educational Psychology, 22, 317-320.