A meta-analysis of the impact of academic and social integration

on persistence of community college students.





Anthony R. Napoli, Ph.D.

Suffolk Community College and

the State University of New York at Stony Brook

and



Paul M. Wortman, Ph.D.

State University of New York at Stony Brook











Running head: META-ANALYSIS





Abstract



Tinto (1975) proposed a prospective model of students persistence which considers a comprehensive set of background and psychosocial factors. Central to the model is the impact social and academic integration have on goal and institutional commitment and the subsequent decision to persist or withdraw from the institution. A number of validation studies, which generally support the model, have been conducted within four-year college and university settings. The few efforts to validate the model on two-year community college have produced a consensus of evidence supporting the importance of academic integration. Support for a social integration-persistence connection has been mixed, however. To further assess the effect of social integration on persistence, and to assess the magnitude of the effect for academic integration, a meta-analysis was conducted. Results indicate that academic integration has significant and beneficial effects on both term-to-term and year-to-year measures of persistence. Social integration was also observed to be significantly and positively linked to term-to-term persistence, but less strongly related to year-to-year persistence.

Over the past decade attrition rates among entering college students have equaled or exceeded completion rates. An American College Testing Program (American College Testing Program) study of a national sample of college entrants (American College Testing Program, 1992), indicates that 50 percent of new full-time four-year college students failed to earn a bachelors degree within five years of entry. At two-year public colleges the graduation rates are substantially lower with less than 39 percent of students completing their associate degree programs within three years of initial entry. Follow-up data on 440,329 1972 high school graduates (National Longitudinal Survey, U.S. Department of Education; 1977, 1983a,b), who enrolled in a two-year college immediately after graduation, indicate only a slight increase to 45 percent in obtaining an associates degree after six years.

Recent ACT data (American College Testing Program 1983, 1986, 1990, 1992) further indicate that a sizeable part of all institutional attrition takes place in the first year. Nationally, rates of first year departure equal 28 and 48 percent for four and two-year public colleges, respectively (American College Testing Program, 1992). An examination of accumulated ACT data (American College Testing Program 1983, 1986, 1990, 1992) indicates that the rate of first year departure has remained extremely stable over the past ten years. This rate has ranged from 28 to 30 percent for four-year public institutions, and from 43 to 44 percent for two-year public colleges.

Tinto's Model of Student Retention.

Tinto (1975, 1987, 1993) proposed a multivariate model of student retention in post secondary institutions to explain student departure from college prior to graduation. His model includes a comprehensive set of demographic, cognitive, psychosocial, and institutional factors. The model proposes that both student characteristics and interactions with the social and academic environments of the institution are the principal determinants of educational goals and institutional commitment. These characteristics and interactions account for the decision to persist or withdraw from college. Working from the end to the beginning of the model, Tinto (1993) proposed that persistence or departure behavior is directly influenced by two dimensions of commitment, namely institutional and goal commitment. Institutional commitment represents the degree to which an individual is motivated to graduate from a specific college or university. Goal commitment, or educational goal commitment, represents the degree to which the individual is committed, or motivated, to earn a college degree in general. In turn, institutional and goal commitments are directly influenced by external commitments or demands and the level of academic and social integration.

Drawing from the theoretical work of Durkheim (1951) and Van Gennep (1960), Tinto (1993) proposed integration as a process in which the individual establishes membership (or fails to establish membership) in the college community. He distinguishes between social integration and academic, or intellectual, integration. The former represents the social ties that result from the day-to-day interactions. The latter results from sharing information, perspectives, and values common to other members of the community.

Integration has also been defined as the extent to which an individual identifies with, or shares and incorporates the normative attitudes and values of his or her instructors and classmates, and becomes a member of the college community (Pascarella and Terenzini, 1991). Satisfying and rewarding interactions with the formal and informal academic and social systems of the institution lead to greater integration and persistence. Unpleasant or limited interactions inhibit integration and decrease the likelihood of persistence. Previous research has shown that academic and social integration are influenced by a variety of factors (Munro, 1981; Pascarella et al., 1983a, 1983b, 1983c, 1985, 1986). These factors include student background characteristics (e.g., age, socioeconomic status, and personality needs), pre-college educational experiences, previous academic achievement, and initial experiences in college.





Empirical Validation of Tinto's Model.

Tinto's model of student retention (1975, 1987, 1993) provides a heuristic and theoretical framework for understanding student behavior. Since its inception, the model has been the subject of many validation efforts. Munro (1981) studied 6,018 first-time, full-time 4-year college entrants. She observed significant direct and indirect effects(1) for academic integration, goal commitment, and high school average on persistence in college. Although the model of student retention postulates an approximate parity between social and academic integration, Munro (1981) found the effect size(2) for the influence of academic integration on persistence to be moderately strong (g = .46, p < .00001). Social integration, however, had no significant direct or indirect effect on persistence. Pascarella and Chapman (1983a, 1983b) also employed a path analytic approach using a large multi-institutional sample of 2,326 college students. They observed that social and academic integration, while not directly influencing persistence, had significant indirect effects on persistence. This influence on persistence operated through institutional commitment and goal commitment.

Another important finding from the Pascarella and Chapman studies (1983a, 1983b) concerned the external validity (Campbell & Stanley, 1966) of the findings across different types of institutional settings. Briefly, cluster-analysis revealed the presence of three distinct institutional groupings: (a) 4-year primary residential institutions, (b) 4-year primary commuter institutions, and (c) 2-year primary commuter institutions. In residential institutions, social integration had both direct and indirect effects on persistence. Conversely, academic integration had no direct or indirect effect on persistence. In both the 4-year and 2-year commuter institutions, however, academic integration indirectly influenced persistence. Among the commuter institutions social integration had neither a direct nor an indirect influence on persistence.

For commuter colleges, the findings for academic integration and social integration have been mixed. In a study of commuter students attending an urban university, direct and indirect effects of academic integration on persistence were observed (Pascarella, Duby, & Iverson, 1983). Contradicting Tinto's formulation, social integration was found to have a negative impact on persistence. That is, an increase in social integration represented an increased risk for withdrawal. Based on these and other findings and earlier results where social integration was found to relate to affiliation needs (Pascarella & Chapman, 1983a), Pascarella and Terenzini (1991) concluded that social integration may be a liability for persistence among commuter students. The investigators argue that commuter institutions are less likely to provide opportunities for social interactions. As a result, person-environment fit problems may occur for individuals with high affiliation needs (denoted by their heightened social integration).

Consistent with these results, academic integration was found to significantly influence social integration and not to influence persistence, among disadvantaged students at a two-year commuter institution (Fox, 1986). Similarly, direct and indirect effects have been observed for academic integration, but not for social integration, among students attending a two-year community college (Mulligan & Hennessy, 1990). Finally, another study failed to detect an effect for social integration among two-year community college students (Halpin, 1990).

Though several investigators have not found support for Tinto's prediction of social integration affecting persistence in commuter samples, others have. Both social and academic integration were found to have direct and indirect effects on long-term (nine-year) persistence and graduation, among two-year community college students (Pascarella et al., 1986). Another study found both integration measures significantly related to persistence in two-year community college students (Bers & Smith, 1991). In fact, this study found that social integration made a larger contribution in discriminating persisters from non-persisters than academic integration. Also, a recent study of two-year community college students, observed strong and nearly identical correlations with persistence for academic integration (r = .51) and social integration (r = .53; Napoli, 1995).

The effect social integration has on persistence is unclear. However, academic integration consistently has a beneficial direct and indirect influence on persistence in two-year commuter schools.

Meta-Analysis of Studies of Academic and Social Integration

As with most narrative literature reviews, subjective appraisals of individual studies provide little guidance in synthesizing meaningful conclusions. This is especially true when the results are mixed as is the case for social integration. Recent advances in meta-analytic procedures (Hedges & Olkin, 1985) or research synthesis (Cooper & Hedges, 1994) provide an objective quantitative alternative for extracting empirically based conclusions by statistically combining results across studies. Therefore, to investigate further the effect of social integration on persistence, in two-year community colleges, a meta-analysis was performed.

The goal of the meta-analysis was to assess the impact and relative importance of social integration and academic integration on persistence/withdrawal behavior of community college students. Based on Tinto's model (1975, 1987, 1993), and the results summarized above, it is expected that academic integration would have a positive influence on persistence. The model also postulates that social interactions between college students and their instructors, and peers, are important mechanisms whereby educational values and goals are exchanged. These interactions, then, might further a student's integration into the educational process.

Although non-residential community college students may have somewhat reduced opportunities for social encounters with classmates (since they do not live on campus) such interaction should still influence persistence decisions. It is, therefore, expected that social integration would have a beneficial effect on persistence among two-year community college students. Specifically, the meta-analysis should show a significant, and positive (i.e., beneficial), influence for both academic and social

integration.



Method

Literature search. Computer generated searches of three major data bases were conducted to identify both published and unpublished articles that assessed the relationship of academic and social integration to persistence among community college students. These data bases were the Educational Resources Information Center (ERIC; U.S. Department of Education, Office of Educational Research and Improvement, Washington, D.C.), PsychINFO (American Psychological Association, Washington, D.C.), and, Dissertation Abstracts Online (University Microfilms International, Ann Arbor, MI). Search parameters were limited to include only English language manuscripts with publication dates extending back to 1980, the year the first validation effort of Tinto's model was conducted by Munro (1980). Keywords for the search were "Persistence" and "Attrition," "Tinto," "Academic Integration," "Social Integration," "Two-year Colleges," and "Community Colleges."



Inclusion criteria. The computer search produced nine published articles and two paper presentations which met the criteria of assessing the impact of academic and social integration on persistence among two-year community college students. From the pool of eleven studies, three (Wolfe, 1993; Mulligan & Hennessy, 1990; Chapman & Pascarella, 1983) did not provide sufficient data to determine the zero order correlation between the integration measures and persistence, a requirement of meta-analysis (Rosenthal, 1991). In addition, two other studies (Munro, 1981; Pascarella & Chapman, 1983) did not report the results for community college samples separately, and so were eliminated from the research synthesis. Finally, one study (Voorhees, 1987) did not assess social integration, and used an unreliable, two-item scale, to assess academic integration, and so was also eliminated from the research synthesis.

From the pool of eleven studies, five were retained for the meta-analysis. Additional data from a recently completed, unpublished study (Napoli, 1995) of 1,011 first-time full-time community college freshmen were also added to the set. Therefore, a final data-set for the meta-analysis consisted of six studies (see in References) with estimable effect sizes for the impact of academic and social integration on persistence.



Coding variables. Effect size (simple correlation coefficients), sample size, and measure of persistence either year-to-year, or term-to-term (discussed below) were extracted from each study. Coding was performed by the first author and an associate. The inter-rater agreement equaled 100 percent for all measures.



Analysis. The effect size estimates (correlation coefficients) for each study, and the mean weighted overall effect size across studies, were expressed as Hedges's g to adjust for sample size bias (Hedges, 1981). Transformation of simple zero-order correlation coefficients to g, and effects size analysis were conducted using DSTAT statistical software (Johnson, 1990). Homogeneity tests were also performed to determine if a single population effect size fit the data adequately, or if moderator variables may be operating across studies producing variable effect sizes. In addition, "fail-safe n" analysis (Rosenthal, 1991) was conducted to assess the "file draw" problem described below.



Results

Academic Integration. As with the studies reviewed earlier, all six of these studies reported significant positive correlations between academic integration and persistence. The effect size estimates for academic integration and their significance tests, as well as the results for the homogeneity test are presented in Table 1. The overall test for homogeneity is significant (Qw (5) = 262.8, p < .001), indicating significant variability in effect sizes among the studies. However, independent tests of significance for each effect size indicate that in all studies academic integration had a significant and positive impact on persistence. Furthermore, the combined overall effect size for academic integration is significant (g = .715, p < .0001). This represents a strong, or large, effect size (Lipsey & Wilson, 1993; Aron & Aron, 1994).



Social Integration. Like the studies reviewed above, the relationship between social integration and persistence was mixed (see Table 2). There is significant heterogeneity of findings among the studies (Qw (5) = 411.7, p < .001). Four out of the six studies had a significant and positive relationship between social integration and persistence (see Table 2). The combined overall mean effect size, obtained by aggregating the weighted effect sizes across the studies, is also significant (g = .459, p .0001) and represents an above average or moderate effect size.

To eliminate any bias resulting from the inclusion of the unpublished study (Napoli, 1995), identified as an outlier study (because it made the greatest contribution to the homogeneity statistic), the meta-analysis was recomputed with this study omitted. Results for the re-analysis remained significant (g = .189, p < .0001). Social integration had a significant and positive impact on persistence. However, the homogeneity statistic also remained significant (Qw = 62.0, p < .0001).



File-draw analysis for Social Integration. One major criticism of meta-analysis is publication bias (Greenhouse & Iyenger, 1994). Publication bias is the increased tendency for journals to publish studies reporting statistically significant findings, and a reduced tendency to publish studies with statistically non-significant results. Publication bias threatens the external validity of meta-analysis (Wortman, 1994). To improve the quality of a research synthesis, Wortman (1994) recommends calculating a "fail-safe n" to estimate the number of "null hypothesis accepting" studies that would be required to reduce the overall effect size to a non-significant level.

Rosenthal (1991) referred to this issue as the "file-drawer" problem (representing the unpublished nonsignificant-result reports residing in the file drawers of researchers). He has developed a statistical procedure to estimate the "fail-safe n" (Rosenthal, 1994). Since the number of published studies identified and available for the present meta-analysis is quite small, publication bias is potentially a threat to the external validity, or "generalizability" of the findings. Therefore, based on the recommendations of Wortman (1994), Begg (1994), Rosenthal (1991), and Becker (1994) a "fail-safe n" analysis was conducted to assess the reliability of the meta-analysis.

Using the effect size estimates from the six studies of social integration, the fail-safe n analysis was preformed employing the procedures developed by Rosenthal(3) (1994). The results of the analysis indicate that approximately 1,396 null-result studies would be required to reduce the combined effect size to a non-significant (p > .05) level. A reexamination of the fail-safe n, excluding the outlier study (Napoli, 1995), indicates that approximately 67 null-result studies would be required to reduce the combined effect size to a non-significant level. Given that a literature search of three major data bases produced only eleven articles that assessed the impact of social integration on retention, among community college students, it seems unlikely that this many studies exist in researchers' file drawers.



Discussion



The meta-analysis results for academic integration indicate that it has a large and positive impact on persistence/withdrawal behavior among community college students. This finding is consistent with Tinto's (1987, 1993) model and research examining this model (Fox, 1986; Mulligan & Hennessy, 1990; Pascarella et al., 1986; Bers & Smith, 1991).

Social integration was also found to play a significant role in persistence/withdrawal decisions. This finding, while consistent with Tinto's (1987, 1993) model, is at odds with his recent conclusions (Tinto, 1993). Tinto (1993) speculates that departure from community college is shaped more by external forces than by internal, campus forces. That is, it is thought that community colleges do not possess sufficient on-campus student communities to foster social integration among students. Results from the present meta-analysis, however, do not support this position. Indeed, the combined overall effect size for social integration was found to be significant, and reflects the important impact social integration has on persistence/withdrawal decisions of community college students. The robustness and reliability of this finding are supported by the results of the fail-safe n analysis.

Nevertheless, the observed heterogeneity of effect sizes across the six studies suggests that the relationship between social integration and persistence may be moderated by some additional factor(s). One possible explanation for the significant variability in effect sizes concerns the measure of persistence employed. Three studies (Bers & Smith, 1991; Napoli, 1995, Axelson & Torres, 1995) assessed term-to-term persistence, whereas, the remaining three studies employed a measure over a longer time interval -- academic year-to-year persistence. Based on these different measures of persistence employed, and the observed differences in effect sizes, it is reasonable to hypothesize that the relationship between social integration and persistence may be moderated by time. That is, as the time between the initial assessment of social integration and persistence increases, the relationship decreases or becomes weaker. There is evidence that reliability decreases over time, which would consequently diminish predictive validity as well (Campbell & Boruch, 1975).

Johnson (1990) has developed a homogeneity test which performs post hoc comparisons of aggregated effect sizes. The results for the comparison of the studies assessing term-to-term persistence and those employing the year-to-year outcome measure are reported in Table 3. Results indicate that the aggregated effect size for the impact of social integration on persistence is significantly greater (2(1) = 64.84, p < .0001) for studies involving term-to-term persistence (g = .641) than those assessing the academic year-to-year outcome (g = .243).

From these findings it appears that, among first-time full-time community college students, social integration has a large impact on more immediate measures of persistence. It also indicates that the influence of one term's level of social integration is moderated, or attenuated, by the length of the persistence interval. In this regard, results from the post hoc analysis indicate that as the length of the time interval increases the impact of social integration on persistence decreases. This finding is not surprising considering the high rates of early (first semester) withdrawal and stop-out behavior (temporary withdrawal) among community college students (Bers & Smith, 1991; Tinto, 1993). Unlike four-year residential institutions, where stop-out behavior is less common and social ties and networks remain relatively stable over several academic semesters or years, social networks within community colleges are less likely to persist over time. Therefore, in community colleges, the ability of one semester's social integration to influence persistence/withdrawal behavior is limited to more short term measures of persistence. Indeed, it is arguable that term-to-term, rather than year-to-year, persistence is a more meaningful measure when studying community college students. This conclusion is consistent with the results of the Bers and Smith (1991) and Napoli (1995) studies, and results reported by Webb (1988) and Romano (1995) who substantially increased the predictive power of their models by studying term-to-term persistence of community college students.

It is also possible that assumptions concerning the relative "unimportance" of social integration in predicting community college student persistence (Tinto, 1993; Pascarella & Terenzini, 1991) may only apply to small schools where opportunities for social integration are limited. In larger two-year institutions social integration may indeed play an important role in persistence/withdrawal decisions. Since information on institutional size was not available in five of the six studies examined above, it was not possible to assess the moderating influence of campus size on the social integration-persistence relationship. It is noteworthy, however, that the Napoli (1995) study, which was conducted at a large community college, found both academic and social integration to have relatively equivalent effects. This finding is consistent with the hypothesis that social integration may operate at large two-year community colleges in a manner similar to that at large four-year colleges.

Several investigators have acknowledged an interplay between institutional size and academic and social integration (Pascarella & Terenzini, 1991; Pascarella & Wolfle, 1985; Tinto, 1993; Chickering & Reisser, 1993). Although these, and other investigators, have consistently treated size as a mediator variable, Tinto points out that:

The effect of size and diversity upon patterns of student departure can be best described as two-pronged. While increased size heightens the possibility that the institution will house a greater variety of social and intellectual communities [and therefore greater opportunity for social integration], it lessens the likelihood that students will have extensive contacts with faculty and staff. The reverse appears to apply to small colleges. Though they tend to be more socially and intellectually homogeneous, they normally provide for greater contact with faculty [academic integration]. (Tinto, 1993, p 80)



Wilson, Gaff, Dienst, Wood, and Bavry (1975) presented evidence suggesting that institutional size is negatively associated with the amount of student-faculty contact. In an eight-institution sample, they found a significant negative relationship between the number of "out of class" student-faculty interactions and institutional size using self-report data obtained from both students and the faculty. It appears then, that observed model differences in social integration that emerge between 4-year residential and 2-year commuter colleges, may also be due to size differences, and not solely institution type.

Within large 2-year community colleges, social integration does occur and has a beneficial impact on persistence (e.g., Napoli, 1995). Therefore, a comprehensive study of the factors related to persistence and attrition among two-year community college students should examine both academic and social integration. The study should also include factors, such as size, which influence social integration and may account for some of the heterogeneity found among the studies.

References



American College Testing Program. (1983). Data compiled from the ACT Institute Data File for 1983. Iowa City, Iowa: American College Testing Program.

American College Testing Program. (1986). Data compiled from the ACT Institute Data File for 1986. Iowa City, Iowa: American College Testing Program.

American College Testing Program. (1990). Data compiled from the ACT Institute Data File for 1990. Iowa City, Iowa: American College Testing Program.

American College Testing Program. (1992). Data compiled from the ACT Institute Data File for 1992. Iowa City, Iowa: American College Testing Program.

American College Testing Program. (1994). The ACT evaluation/service for educational institutions and agencies: Users Guide (6th ed.). Iowa City, Iowa, ACT.

Aron, A., and Aron, E. (1994). Statistics for psychology. Englewood Cliffs, NJ: Prentice Hall.

Axelson, R., and Torres, D. (1995). Modeling first-semester integration and departure of community college students. Paper presented at the Annual Forum of the Association for Institutional Research, Boston.

Becker, J.B. (1994). Combining significance levels Bias. In Cooper, H., & Hedges, L.V. (Eds.). (1994). The Handbook of research synthesis (pp. 215-230). New York: Russell Sage Foundation.

Begg, C.B. (1994). Publication Bias. In Cooper, H., & Hedges, L.V. (Eds.). (1994). The Handbook of research synthesis (pp. 399-410). New York: Russell Sage Foundation.

Bers, T.H., and Smith, K.E. (1991). Persistence of community college students: The influence of student intent and academic and social integration. Research in Higher Education 32(5), 539-556.

Campbell, D.T., & Stanley, J.C. (1966). Experimental and quasi-experimental designs for research. Boston: Houghton Mifflin.

Campbell, D.T., & Boruch R.F. (1975). Making the case for randomized assignment to treatments by considering the Alternatives: Six ways in which quasi-experimental evaluations in compensatory education tend to underestimate effects. In Bennett, C.A., & Lumsdaine, A.A. (Eds.) (1975). Evaluation and Experiment: Some critical issues in assessing social programs. New York: Academic Press.

Chapman, D.W., and Pascarella, E.T. (1983). Predictors of academic and social integration of college students. Research in Higher Education, 19(3), 295-322.

Chickering, A.W. (1974). Commuting verses resident

students. San Francisco: Jossey-Bass.

Cooper, H., & Hedges, L.V. (Eds.). (1994). The Handbook of research synthesis. New York: Russell Sage Foundation.

Fox, R. (1986). Application of a conceptual model of college withdrawal to disadvantaged students. American Educational Research Journal, 23, 415-424.

Greenhouse, J.B., & Iyenger, S. (1994) Sensitivity analysis and diagnosis. In Cooper, H., & Hedges, L.V. (Eds.). (1994). The Handbook of research synthesis (pp. 383-389). New York: Russell Sage Foundation.

Halpin, R.L. (1990). An application of the Tinto model to the analysis of freshmen persistence in a community college. Community college review, 17(4), 22-32.

Hedges, L.V., (1981). Distribution theory for Glass's estimate of effect size and related estimates. Journal of Educational Statistics, 6, 107-128.

Hedges, L.V., & Olkin, I. (1985). Statistical methods for meta-analysis. New York: Academic Press.

Johnson, B.T. (1990) DSTAT: Software for the meta-analytic review of research literatures. Hillsdale, NJ: Erlbaum.

Lipsey, M.W., & Wilson, D.B. (1993). The efficacy of psychological, educational, and behavioral treatments: confirmation from meta-analysis. American Psychologist 48(12) , 1181-1209.

Mulligan, S.C., and Hennessy, J.J., (1990). Persistence in a community college: Testing attrition models. Paper presented at the annual meeting of the American Educational Research Association, Boston.

Munro, B. (1981). Dropouts from higher education: Path analysis of a national sample. American Educational Research Journal, 81, 133-41.

Napoli A. (1995). An examination of the relative importance of academic and social integration among community college students. Unpublished manuscript, Suffolk Community College.

Pascarella, E.T., and Chapman, D. (1983a). A multi-institutional, path analytic validation of Tinto's model of college withdrawal. American Educational Research Journal 20(1), 87-102.

Pascarella, E.T., and Chapman, D. (1983b). Validation of a theoretical model of college withdrawal: Interactive effects in a multi-institutional sample. Research in Higher Education 19(1), 25-48.

Pascarella, E.T., and Terenzini, P.T. (1983). Predicting voluntary freshmen year persistence/withdrawal behavior in a residential university: A path analytic validation of Tinto's model. Journal of Educational Psychology, 75, 215-216.

Pascarella. E.T., Duby, P.B., and Iverson, B. (1983). A test and re-conceptualization of a theoretical model of college withdrawal in a commuter institution setting. Sociology of Education, 56, 88-100.

Pascarella, E.T., and Wolfle, L.M. (1985). Persistence in higher education: A nine-year test of a theoretical model. Paper presented at the annual meeting of the American Educational Research Association, Chicago.

Pascarella, E.T., Smart, J.C., and Ethington, C.A. (1986). Long-term persistence of two-year college students. Research in Higher Education, 24(1) 47-71.

Pascarella, E.T., and Terenzini, P.T. (1991). How college affects students: Findings and insights from twenty years of research. San Francisco: Jossey-Bass.

Romano, R.M. (1995). First-year attrition and retention at a community college. Journal of Applied Research in the Community College, 2(2), 169-177.

Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.

Rosenthal, R. (1991). Meta-analytic procedures for social research. Beverly Hills, CA: Sage.

Rosenthal, R. (1994). Parametric measures of effect size. In Cooper, H., & Hedges, L.V. (Eds.). (1994). The Handbook of research synthesis (pp. 231-244). New York: Russell Sage Foundation.

Tinto, V. (1975). Dropout from higher education: A theoretical synthesis of recent research. Review of Educational Research 45, 89-125.

Tinto, V. (1987). Leaving College: Rethinking the causes and cures of student attrition. Chicago: University of Chicago Press.

Tinto, V. (1993). Leaving College: Rethinking the causes and cures of student attrition (2nd ed.). Chicago: University of Chicago Press.

U.S. Department of Education, National Center for Educational Statistics (1977). Withdrawal from institutions of higher education. Washington, D.C.: U.S. Government Printing Office.

U.S. Department of Education, National Center for Educational Statistics (1983a). Digest of Educational Statistics. Washington, D.C.: U.S. Government Printing Office.

U.S. Department of Education, National Center for Educational Statistics (1983b). Participation of black students in higher education: A statistical profile from 1970-71 to 1980-81. Washington, D.C.: U.S. Government Printing Office.

Voorhees, R.A. (1987). Towards building models of community college persistence: A logit analysis. Research in Higher Education, 26(2): 115-129.

Webb, M.W. (1988). Freshman year retention at three campuses of a large urban community college district: 1983-1986. Community and Junior College Quarterly of Research and Practice, 12, 213-242.

Wilson, R., Gaff, J., Dienst, R., Wood, L., and Bavry, J. (1975). College professors and their impact on students. New York: Wiley-Interscience.

Wolfe, J.S. (1993). Institutional integration, academic success, and persistence of first-year commuter and resident students. Journal of College Student Development, 34, 321-326.

Wortman, P.M., (1994). Judging research quality. In Cooper, H., & Hedges, L.V. (Eds.) 1994. The Handbook of research synthesis (pp. 97-110). New York: Russell Sage Foundation.



Study included in meta-analysis.

Table 1



Analysis of the Effect Sizes for the Impact of Academic Integration on the Persistence of Community College Students.



Study g r p.
Pascarella et al. (1983) 0.3860 .1900 .0001
Fox (1986) 1.0924 .4800 .0001
Pascarella et al. (1986) 0.7224 .3400 .0001
Bers & Smith (1991) 0.5079 .2470 .0001
Napoli (1995) 1.1849 .5100 .0001
Axelson & Torres (1995) 0.1402 .0700 .0070
Overall 0.7147 .3365 .0001

Qw(5)= 262.813; p = 0.0000;

Note: Effect sizes are corrected for sample size bias.



Table 2

Analysis of the Effect Sizes for the Impact of Social Integration on the Persistence of Community College Students.

Study g r p.
Pascarella et al. (1983) -0.1802 -.0900 .0367
Fox (1986) 0.1200 .0600 .0768
Pascarella et al. (1986) 0.4506 .2200 .0000
Bers & Smith (1991) 0.4013 .1974 .0001
Napoli (1995) 1.2425 .5280 .0000
Axelson & Torres (1995) 0.2000 .0100 .7002
Overall 0.4594 .2239 .0000

Qw(5)= 411.69; p = .0001

Largest outlier is Napoli



Table 3

Analysis of the Moderating Influence of Persistence Measure (term-to-term verses academic year-to-year) on relationship between Social Integration and Persistence.

Persistence Measure k g r p.
Term-to-term 3 .641 .121 .001
Year-to-year 3 .243 .305 .001

Note: k = number of studies.

2 (1) = 64.84, p. < .0001

1. The term direct effect represents the unmediated influence of one variable on another. It occurs when the impact of one variable on another is direct and does not pass through a third, or intervening, variable. An indirect effect is one in which the influence of one variable on another is mediated by, or passes through, another variable. For example, in a three variable model, the effect of variable A (e.g., life stress) on variable C (e.g., illness) is mediated, or passes through, variable B (e.g., negligent health practices).

2. Hedges' g for r adjusts for sample size bias. It is computed as: g = (2r/1-r2) . (N-2)/N (see Hedges, 1981).

3. To calculate the "fail-safe n" Rosenthal's formula was employed as follows:



X = (Z)2 - K

Zcv2



where X is the fail-safe N, K is the number of studies included in the meta-analysis, Z, is the sum of Z obtained for K studies, and Zcv is the critical value of the test statistic at a given alpha level.