Abstract

Since the late 1970s, researchers have examined the relationship between school building condition and student performance. Though many literature reviews have claimed that a relationship exists, no meta-analysis has quantitatively examined this literature. The purpose of this review was to synthesize the existing literature on the relationship between building condition and student performance. Means for the semi-partial ( = 0.10) and bivariate ( = 0.12) correlations were relatively small but significantly different, supporting the claim that school building condition is related to student performance. Furthermore, results revealed that the magnitude of the correlation varied as a function of a number of moderator variables. For instance, the building condition feature measured, instrument type, subject area measured, and grade level affect the association between school building condition and student performance. Our findings offer useful information for educational leaders, policy makers, and researchers.

1.  Introduction

For many years researchers have examined the relationship between school building condition and student performance. The idea of this relationship has been driven by the fact that many educational researchers believe that the learning environment can influence how well students learn and teachers teach (Cash 1993; Hines 1996; Earthman and Lemasters 2011). Because many of our teachers are teaching (and students are educated) in environments that prohibit a profound learning experience, different efforts have been taken to reveal the condition of our nation's schools.

In 2006, a documentary entitled Corridor of Shame (Ferillo 2006) reported on the inadequate funding for several schools in the state of South Carolina. The documentary presented the deplorable and dangerous conditions of schools in which many low-performing students were forced to learn. In addition, Schneider (2003) surveyed a large sample of Chicago and Washington, DC, teachers to identify problems faced within their school facilities. Over 44 percent of Chicago teachers and 68 percent of Washington teachers reported that their school facilities were too noisy. This is an issue that can clearly make it difficult for students to concentrate, as well as difficult for teachers to teach.

A primary goal of the American Federation of Teachers (AFT) is to improve the condition of school facilities by enforcing higher standards and accountability to educational leaders. The AFT believes that educational leaders have a responsibility to ensure all students are educated in environments that are healthy, safe, clean, and well-maintained. Unfortunately, to date this is not the case. According to the U.S. Department of Education, during the 2012 fiscal year, approximately 53 percent of K–12 public school buildings were in need of modernizations, renovations, and repairs (USDOE 2014). Given the decline in building conditions and the increased efforts to improve student performance, the influence of school building conditions on student performance is of interest to many educators, researchers, and policy makers.

Since the late 1970s, a number of studies have examined the relationship between building condition and student performance. Although the idea that the condition of the school buildings can affect student achievement seems coherent, the findings in this area have been inconsistent. Generally speaking, some empirical studies have reported a positive relationship between building condition and student achievement (e.g., Berner 1993; Al-Enezi 2002; Duran-Nurucki 2008), while other studies have indicated that no relationship exists (e.g., Cervantes 1999; Guy 2001; Picus et al. 2005).

Over the years many researchers have sought to quantify the literature on building condition and student performance (Weinstein 1979; McGuffey 1982; Lemasters 1997; Earthman 2004; Bailey 2009; Stewart 2010). Because of the variation seen across studies in the measures representing the conditions of school facilities, however, as well as the study designs and methods, researchers have taken a narrative approach to summarize this literature.

Consider first the most recent extensive review by Stewart (2010). Stewart focused on the effect of four dimensions of the school's physical environment (building condition, building age, artificial lighting, and natural lighting) on student learning. The review compiled and examined 42 quantitative studies. Of the 42 studies, 16 included analyses examining the relationship between building condition and student performance. His synthesis revealed that 50 percent of the analyses reported a positive relationship, 38 percent reported no relationship, and 12 percent reported an inverse relationship between building condition and student performance. On the basis of this ”vote count” of results, Stewart concluded that there is a weak association between building conditions and student learning.

Bailey (2009) focused on the effect of building condition on student and teacher performance, health, behavior, attitude, and absenteeism. The review included 54 studies from 1998 to 2008. Of the 54 studies, 18 examined the relationship between building condition and student performance. Bailey made no statistical attempt to combine results across studies. Nonetheless, Bailey concluded that the majority of studies showed that the condition of school facility has a direct influence on student performance. In addition, Lemaster (1997) synthesized 53 studies from 1987 to 1997 on the relationship between school facilities, student achievement, and student behavior. Results indicated that student achievement and behavior were directly related to the condition of school facilities.

A major inconsistency in the literature is the measures used to represent the building conditions. The school facility is a multi-dimensional construct of structural features (e.g., building age, heating and air systems) and cosmetic features (e.g., interior and external painting). Some researchers have assessed individual features of the school facility. Others have utilized instruments such as the Commonwealth Assessment of Physical Environment (CAPE), developed by Cash (1993), to obtain a composite score of the structural and cosmetic components. Although there is no simple or correct way to define or measure the building condition, the inclusion of extraneous features probably led to the mixed findings seen across studies.

In addition, different indicators of academic achievement have been used. The most commonly used indicator of student achievement relates to standardized test scores in specific academic areas (e.g., science). Other indicators of academic achievement, however, such as school grade point average, passing percentage, and attainment may have been used as educational outcomes. Although there is no correct way of assessing academic achievement, the difference in the measurement scales may have also contributed to the inconsistent findings.

In an effort to strengthen and guide future research in this area, Earthman and Lemasters (2011) developed a theoretical framework that attempts to explain how the school building conditions influence students and teachers. Their theoretical model focused on six propositions: (1) the role of educational leadership and financial ability; (2) the role of maintenance and operational staff; (3) the influence of building conditions on the attitudes of parents, students, and teachers; (4) the impact of building conditions on students’ self-worth; (5) the impact of students’ reflection of building conditions on student achievement; and (6) the impact of building conditions on student performance. The authors attempt to explain why the conditions of school buildings are important, and the mechanisms through which the building conditions can influence students’ achievement and behavior. The authors claim there exists a positive relationship between building condition and student performance, with the most positive results seen in the math and science subject areas.

Although many researchers have reviewed this literature, with many claiming that building condition and student performance are related (Lemasters 1997; Bailey 2009; Stewart 2010), the extent of the relationship remains unclear. In addition, vote counting was used to make and support this claim. But vote counting has been shown to be a flawed procedure for research synthesis (Hedges and Olkin 1985). Vote counting assigns equal weights to each study regardless of its precision and sample size. This may lead to bias and misleading results; thus, results should be interpreted with caution (Hedges and Olkin 1985).

The goal of this review is to provide a better understanding of the overall relationship between building condition and student performance, using a meta-analytic approach. Meta-analysis is a quantitative method of research synthesis that is used to combine results across studies. Meta-analysis offers great advantage over the techniques used in prior reviews. First, meta-analysis allows estimation of the strength of the relationship. Second, unlike vote counting, meta-analysis accounts for unequal precision by weighting each study by the inverse of its variance (Hedges and Olkin 1985). Moreover, because findings have been mixed, this technique allows an examination of the influence of moderators on the relationship. This will effectively address concerns relating to contextual and methodological characteristics, which in turn can provide useful information for educational leaders and guide future research. Thus we addressed two general questions:

  1. What is the overall strength of the relationship between building condition and academic achievement?

  2. What study characteristics moderate the relationship between building condition and academic achievement?

In particular, we asked the following nine questions:

(1) Does the building-condition feature measured affect the association between school building and students’ academic achievement? Measures of building condition can be represented by a single component for cosmetic and/or structural building conditions or a combined score that reflects both components. We were interested in whether the correlation between school building condition and students’ academic achievement depends on the building condition measure used. In order to understand which features relate most strongly to student outcome, we examined the relationship between student performance and individual features of the school facility. It is possible that certain features of the school facility have more of an effect than other features on student achievement. Knowing such information can provide insight for researchers interested in this area, as well as educational leaders and policy makers.

(2) Does the academic achievement measure used affect the association between school building condition and students’ academic achievement? Earthman and Lemasters (2011) stated, “The most positive results have been noted for the subject areas of mathematics and science” (p. 30). Thus we investigated whether the relationship between school building condition and students’ academic achievement depends on the academic area measured. Picus et al. (2005) stated that the “more puzzling issue is why some subjects, like math and social studies, would be affected differently” (p. 79). Perhaps higher effects are seen in studies that examine the relationship for science achievement than studies that used language arts as the outcome. Because science is a subject that requires equipment and supplies for hands-on experiences, poorly kept schools may lack adequate science facilities, resulting in a higher impact of building condition.

(3) Does the specific instrument type used to measure building condition affect the association between school building condition and students’ academic achievement? Another source of variability in the relationship between school building condition and students’ academic achievement may stem from the instrument used to assess the condition of building. Bailey (2009) noted, “There were more instances of significance in studies that incorporated the CAPE” (p. 191). Thus we investigated whether the magnitude of the relationship varies between studies using the CAPE versus studies using other instruments.

(4) Does the rater who assessed building conditions affect the association between school building condition and students’ academic achievement? We investigated whether the correlation between school building condition and students’ academic achievement depends on the rater who assessed the building condition. We expect to see a higher effect in studies evaluated by school personnel than those assessed by external personnel. In general, when making judgments or decisions, people tend to place more weight on the negative aspects than the positive aspects (Kahneman and Tversky 1984). One can assume that when teachers rate conditions, they likely consider how the conditions have influenced their own students, which may lead to a bias estimate in an upward direction.

(5) Does the grade level of the students contribute to differences in the relationship between school building condition and students’ academic achievement? We are also interested in whether the correlation between school building condition and students’ academic achievement depends on the grade level of the students. According to McGuffey (1982), “facilities may have a differential impact on the performance of pupils in different grades and for different subjects” (p. 276). We believe a stronger effect will be seen for younger students than for students who are older. Younger children undergo different cognitive and psychological developmental processes (Wellman and Lagattuta 2004). A noisy or uncomfortable environment may affect their ability to acquire the skills intended during the primary school years, and hence moderating the correlation between building condition and academic performance.

(6) Does study year contribute to the variation in the relationship between school building condition and students’ academic achievement? We test whether the relationship between school building condition and students’ academic achievement has changed over time. Although we do not expect to see any changes as it relates to the year of publication, we assessed this to rule out potential confounding variables.

(7) Does publication status affect the association between school building condition and students’ academic achievement? Variation in the correlation between school building condition and students’ academic achievement may stem from publication type. We will explore whether the connection differs for published and unpublished documents, and will see if variation exists within each type of document. If the relationship is significantly higher in published documents, this could be evidence of publication bias (Rothstein, Sutton, and Borenstein 2005).

(8) Does controlling for socioeconomic status (SES) affect the association between school building condition and students’ academic achievement? Research suggests SES accounts for a large percent of the variance explained in student achievement (Goldhaber and Brewer 1997; Sirin 2005). In some cases, researchers have examined this relationship using more sophisticated analyses to control for SES (e.g., Berner 1993; Duran-Nurucki 2008; Evans, Sipple, and Yoo 2010). Thus, we will explore whether the relationship between school building condition and students’ academic achievement varies as a function of whether or not researchers statistically controlled for SES. Not controlling for this variable may bias the estimates upward, making the relationship between building condition and student achievement higher.

(9) Does controlling for attendance affect the association between school building condition and students’ academic achievement? Variation in the size of the correlation may also stem from whether or not researchers statistically controlled for attendance. Research shows students who attend school on a regular basis have higher achievement scores than their peers who attend school less frequently (Lamdin 1996). Failure to control for this variable may bias the estimates upward, impacting the conclusions drawn.

2.  Data

Literature Search

Seven databases and one Web site were used in our search for articles and dissertations: ERIC, EBSCO Host, ISI, Science Direct, ProQuest, JSTOR, Dissertation Abstracts, and NCEF (National Clearinghouse for Educational Facilities). The keywords used in our search were building condition, building quality or school facilities AND academic achievement or academic performance. A review of abstracts for the initial search produced 623 papers, with 215 overlapping studies. In addition we checked the reference lists of Earthman and Lemasters (2011), Stewart (2010), and Bailey (2009) to obtain studies from the prior reviews.

Inclusion Criteria

The studies that were selected for review (1) targeted K–12 students, (2) focused on the condition of the school's overall building condition or physical components of cosmetic or structural features, (3) targeted student achievement, (4) measured student achievement in terms of standardized test scores, (5) examined the relationship using ordinary least squares multiple regression analyses or via the bivariate correlation, and (6) reported the Pearson moment correlation or the semi-partial correlation coefficient, or appropriate statistics (i.e., correlations, R2, t statistics) that can be converted to an effect size by using appropriate statistical transformation formulas.

Exclusion Criteria

The studies that were omitted from review (1) focused on building designs and academic achievement, (2) measured student outcomes in terms of passing percentage, (3) measured student outcomes in terms of school dropout, (4) performed comparison-group design analyses, and (5) did not report the Pearson moment correlation or appropriate statistics to compute the semi-partial correlation coefficient.

Coding Procedures

Because the findings in this area have been mixed and the knowledge in this area is thin, we examined the influence of study characteristics on the strength of the relationship. Information on the building condition measure (cosmetic, structural, composite), academic outcome (math, language arts, science, other, or combined), building instrument type (CAPE or other instrument), rater who assessed building-condition (school personnel or external personnel such as subcontractor, construction personnel, or architect), grade level (elementary, middle, high school, or combined), publication source (dissertation or journal article), and study year was coded for each study. In addition, for studies reporting multiple regression analyses, we coded whether or not regression models controlled for socioeconomic status and student attendance.

In some cases, researchers reported multiple correlations for the same or different samples on different building feature measures (e.g., Al-Enezi 2002; O’Sullivan 2006). For these studies, individual features were coded as a cosmetic or structural dimension. In particular, features relating to the age of the building, noise, heating and air systems, maintenance, flooring, lighting, and condition of science labs were classified as structural features, whereas comfort features like internal painting, external painting, location of graffiti, or cleanliness were coded as cosmetic features. For studies reporting the correlation for individual features and an overall building condition measure, only the independently assessed features were used to reduce common method bias. Individual components were then categorized into subgroups relating to temperature, physical appearance, noise, age, or a combination thereof to identify which features show the largest effect.

Coding was completed by the first and second author. Reliability among coders was calculated to determine the level of agreement between the coders for each variable. The overall mean level of agreement for study coding characteristics with multiple regressions was 92.63 percent, and 86.67 percent for studies reporting Pearson moment correlation. All discrepancies were resolved by reexamining the data. The data included in our review consist of 594 effect measures from the nine studies reporting correlational analyses and 36 effect-size measures from the nine studies reporting regression analyses.

3.  Meta-Analysis Procedures

Effect Sizes

Using meta-analysis, this research seeks to estimate the average relationship across studies. The correlation r and the semi-partial correlation (Aloe and Becker 2012) indexes were used to represent the magnitude of association between building condition and academic achievement.

Bivariate Correlations

Studies reporting the relationship via bivariate correlations, typically reported as Pearson's r, represent the linear relationship between building condition and student performance. The Fisher's z transformation in equation 1 was used to convert the correlations to normally distributed values and stabilize the variance (Fisher 1924). For the correlation from study i we obtained
formula
1
After running the analysis, equation 2 was used to convert results back to the original correlation scale:
formula
2

Semi-partial Correlations

The semi-partial correlation () was used for studies examining the effect of building condition on student performance via multiple regressions. Provided that sufficient information is reported within studies, we can obtain the semi-partial correlation effect size estimate. The represents the correlation between building condition and academic performance after the variance between building condition and other predictor variables in the same model have been removed. Aloe and Becker (2011) examined the behavior of this index under certain conditions. When no correlation exists among the predictor variables in the regression model, the semi-partial correlation will yield similar results to the zero-order correlation (Aloe and Becker 2012). If multicollinearity exists, however, the value will often be lower than the value. The value of was computed using equation 3 (Aloe and Becker 2012), as
formula
3
where is the sample size for study i, is the number of predictors in the model in study i, is the t statistic for study i for the regression coefficient , and is the squared multiple correlation between all predictors in the model and the outcome variable for study i.

Analyses

Under the fixed-effect model, the Q statistic was calculated to analyze the heterogeneity among studies for each effect size index ( or ) (Hedges and Olkin 1985). The homogeneity assesses whether all the studies arise from one population. Under the fixed-effect model, effect sizes were weighted using the inverse-variance , with , where is the sample size. If the value of the Q statistic exceeds the critical value of the χ2 distribution with degrees of freedom, homogeneity of effect size is rejected (Hedges and Olkin 1985). In such cases, the effect-size estimates do not share the same population effect size under the fixed-effect model, and the random-effects model should be selected to obtain the overall index mean estimate.

Moderator Analysis

Next, regression analysis and weighted analyses of variance were used to examine the influence of moderator variables. For each effect size ( or ), regression analysis was used to examine the continuous predictor—study year. Analysis of variance–like categorical analyses were conducted for each effect size ( or ) and all coded variables except study year. The weights were estimated based on the mixed-effects model. The mixed-effects model incorporates components of the fixed-effect and random-effects model to assess heterogeneity between studies (see, e.g., Hedges and Pigott 2004).

4.  Results

The results are presented in two sections. The first part of our analysis is for studies reporting bivariate correlations. The second part includes studies reporting analyses with multiple regression models and synthesized using rsp. Both sections report the overall magnitude of the relationship and the results of the moderator analyses.

Correlations

Under the random-effects model, the weighted overall mean correlation was 0.12 (SE = 0.01), a 95 percent confidence interval from 0.11 to 0.14. The confidence intervals for 594 correlations, ranging from –0.56 to 0.71, are illustrated in figure 1, where the horizontal solid line represents the weighted mean effect. The homogeneity test revealed significant results, Q(593) = 1,166.59, p < 0.001, indicating that the correlations did not come from the same population. Thus, on average, school building condition is positively related to academic achievement to a weak degree.

Figure 1.

Confidence Intervals for the Correlations.

Figure 1.

Confidence Intervals for the Correlations.

Next, the results of the categorical analyses are reported in table 1 for study characteristics and table 2 for building-condition characteristics. The tables include the number of effects and the results of the mixed-effect weighted mean, confidence interval, and Q statistics. The results from regression analyses on the predictor variable study year are given.

Table 1. 
Analyses of Correlations by Study Characteristics
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Outcome measure      13.65 (4)* 
Mathematics 92 0.14 0.10 0.18 197.00 (91)**  
Language Arts 193 0.14 0.11 0.17 396.06 (192)**  
Science 92 0.17 0.11 0.24 244.34 (91)**  
Other 212 0.08 0.05 0.11 282.25 (211)**  
Composite 0.12 −0.03 0.27 4.57 (4)  
School level      47.02 (3)** 
Elementary 61 0.20 0.16 0.25 106.76 (60)**  
Middle 60 0.26 0.21 0.31 79.51 (59)*  
High 467 0.10 0.07 0.12 891.82 (466)**  
Elementary, Middle, −0.21 −0.52 0.15 2.96 (5)  
and High 
Publication status      28.15 (1)** 
Journal 134 0.21 0.17 0.24 229.19 (133)**  
Dissertation 460 0.10 0.08 0.12 888.13 (459)**  
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Outcome measure      13.65 (4)* 
Mathematics 92 0.14 0.10 0.18 197.00 (91)**  
Language Arts 193 0.14 0.11 0.17 396.06 (192)**  
Science 92 0.17 0.11 0.24 244.34 (91)**  
Other 212 0.08 0.05 0.11 282.25 (211)**  
Composite 0.12 −0.03 0.27 4.57 (4)  
School level      47.02 (3)** 
Elementary 61 0.20 0.16 0.25 106.76 (60)**  
Middle 60 0.26 0.21 0.31 79.51 (59)*  
High 467 0.10 0.07 0.12 891.82 (466)**  
Elementary, Middle, −0.21 −0.52 0.15 2.96 (5)  
and High 
Publication status      28.15 (1)** 
Journal 134 0.21 0.17 0.24 229.19 (133)**  
Dissertation 460 0.10 0.08 0.12 888.13 (459)**  

Notes:aQw is based on the fixed-effect model.

bQB is based on the mixed-effect model.

*Significant at α = 0.05; **significant at α = 0.001.

Table 2. 
Analyses of Correlations by Building Condition Characteristics
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Building condition measure      41.29 (2)** 
Cosmetic 279 0.10 0.07 0.12 428.61 (278)**  
Structural 285 0.17 0.14 0.20 680.01 (284)**  
Overall 30 0.00 −0.04 0.05 11.00 (29)  
Individual features      122.18 (5)** 
Temperature 0.10 0.00 0.20 13.52 (5)*  
Physical 343 0.11 0.09 0.13 567.18 (342)**  
Electrical −0.02 −0.14 0.10 18.46 (5)*  
Noise 135 0.27 0.24 0.29 136.94 (134)  
Age 66 0.01 −0.07 0.08 211.24 (65)**  
Combination 38 0.04 −0.01 0.08 44.33 (37)  
Rater      25.21 (1)** 
Internal personnel 455 0.10 0.08 0.12 884.63 (454)**  
External personnel 139 0.20 0.17 0.24 236.25 (138)**  
Building instrument type      25.91 (1)** 
CAPE 454 0.10 0.08 0.12 882.51 (453)**  
Others 140 0.20 0.17 0.24 236.70 (139)**  
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Building condition measure      41.29 (2)** 
Cosmetic 279 0.10 0.07 0.12 428.61 (278)**  
Structural 285 0.17 0.14 0.20 680.01 (284)**  
Overall 30 0.00 −0.04 0.05 11.00 (29)  
Individual features      122.18 (5)** 
Temperature 0.10 0.00 0.20 13.52 (5)*  
Physical 343 0.11 0.09 0.13 567.18 (342)**  
Electrical −0.02 −0.14 0.10 18.46 (5)*  
Noise 135 0.27 0.24 0.29 136.94 (134)  
Age 66 0.01 −0.07 0.08 211.24 (65)**  
Combination 38 0.04 −0.01 0.08 44.33 (37)  
Rater      25.21 (1)** 
Internal personnel 455 0.10 0.08 0.12 884.63 (454)**  
External personnel 139 0.20 0.17 0.24 236.25 (138)**  
Building instrument type      25.91 (1)** 
CAPE 454 0.10 0.08 0.12 882.51 (453)**  
Others 140 0.20 0.17 0.24 236.70 (139)**  

Notes:aQw is based on the fixed-effect model.

bQB is based on the mixed-effect model.

*Significant at α = 0.05; **significant at α = 0.001.

Academic Measure

Our data for composite scores include one effect size combining English and math scores; and four effect sizes including composite test scores in core subjects (i.e., English, math, and science) and foundation subjects (physical education, arts, humanities, etc.). Results were statistically significant, QB (4) = 13.65, p < 0.05, indicating that the correlation between building condition and student performance varied as a function of the outcome measure. Outcome based on “science” measure produced the highest mean effect (0.17), followed by language arts (0.14), math (0.14), and composite scores (0.12). The measure reporting student achievement using “other” indicator was significantly lower (0.08). In addition, significant variation exists within the areas of science, language arts, other, and math scores.

School Level

Four different categories of school levels were examined: elementary school, middle school, high school, and combined schools (elementary school, middle school, and high school). The means by school level were significantly different from zero, (QB (3) = 47.02, p < 0.001), implying that this relationship differs as a function of school levels. The mean effect sizes (ESs) for elementary and middle-school levels were significantly higher than the mean effect ESs for high school level. In addition, significant variation exists within the school levels.

Publication Status

Significant variation, (QB (1) = 28.15, p < 0.001), was seen in the analysis of publication status. Published documents had a significantly higher mean correlation (0.21) compared with unpublished documents (0.10), indicating the possible presence of publication bias.

Measure of Building Condition

The building condition category included cosmetic, structural, and overall measures. The groups were statistically different from zero, (QB (2) = 41.29, p < 0.001), indicating the measure representing building condition can explain some of the variability in the correlation. The mean ESs for structural and cosmetic features were higher compared with the mean effect of the overall/composite measure. In addition, the studies using cosmetic and structural measures were heterogeneous, suggesting significant variation exists among the correlations of the cosmetic and structural building features. Further exploration showed that noise level had the highest mean ES (0.27), followed by features representing physical appearance (e.g., wall painting and graffiti) (0.11).

Rater Who Assessed Building Condition

The rater who assessed the building condition included mean correlations for internal and external personnel. The test of between-group differences was statistically significant, with QB (1) = 25.21 (p < 0.001). Post hoc comparison revealed that studies rated by external personnel (0.20) had a higher mean effect than studies rated by internal personnel (0.10). In addition, within effects for external and internal personnel correlations were heterogeneous, indicating that effects varied significantly.

Instrument Type

The analysis of instrument type used to evaluate building condition was significant, (QB (1) = 25.91, p < 0.001), indicating that the instrument type explained some of the variability in this relationship. Higher effects were seen in “other” instruments (0.20) than those using the CAPE survey (0.10). In addition, significant variations exist within both instruments used to assess the building conditions.

Study Year

The year of study was examined to determine its predictive power for the correlation. Study year significantly related to the magnitude of this relationship, under the mixed-effects model, Q(1) = 3.99, p < 0.05 (b = 0.0071, SE = 0.0037).

Semi-partial Correlations (rsp Values)

The second portion of our analysis examined the effect of the building condition on student achievement in studies reporting multiple regressions. The overall estimated mean was = 0.10 (SE = 0.01), with a 95 percent confidence interval from 0.07 to 0.12. The forest plot of the confidence intervals for thirty-six semi-partial correlation values, which ranged from –0.30 to 0.30, can be seen in figure 2. The homogeneity test revealed significant results (Q(35) = 84.60, p < 0.001), indicating the values did not come from the same population. Therefore, on average, the overall building condition is again related to student performance, if only weakly.

Figure 2.

Confidence Intervals for the Semi-partial Correlations.

Figure 2.

Confidence Intervals for the Semi-partial Correlations.

Categorical analyses are reported under the mixed-effects model for study characteristics, shown in table 3, building condition characteristics, shown in table 4, and methodological characteristics, shown in table 5. These tables include the number of effects, weighted mean, confidence interval, and Q statistics. Results for the predictor variable study year follow.

Table 3. 
Analyses of Semi-partial Correlations by Study Characteristics
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Outcome measure      8.56 (4) 
Mathematics 11 0.14 0.11 0.17 9.64 (10)  
Language Arts 11 0.10 0.04 0.15 39.33 (10)**  
Science 0.06 0.00 0.12 1.94 (2)  
Other 0.08 0.04 0.13 8.64 (6)  
Composite 0.01 −0.16 0.18 10.04 (3)*  
School levelc      12.52 (2)* 
Elementary 29 0.11 0.08 0.13 60.48 (28)**  
High −0.21 −0.38 −0.03 0.48 (3)  
All 0.09 −0.33 0.51 6.77 (1)*  
Publication status      2.83 (1) 
Journal 16 0.08 0.03 0.12 49.41 (15)**  
Dissertation 20 0.12 0.09 0.16 23.79 (19)  
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Outcome measure      8.56 (4) 
Mathematics 11 0.14 0.11 0.17 9.64 (10)  
Language Arts 11 0.10 0.04 0.15 39.33 (10)**  
Science 0.06 0.00 0.12 1.94 (2)  
Other 0.08 0.04 0.13 8.64 (6)  
Composite 0.01 −0.16 0.18 10.04 (3)*  
School levelc      12.52 (2)* 
Elementary 29 0.11 0.08 0.13 60.48 (28)**  
High −0.21 −0.38 −0.03 0.48 (3)  
All 0.09 −0.33 0.51 6.77 (1)*  
Publication status      2.83 (1) 
Journal 16 0.08 0.03 0.12 49.41 (15)**  
Dissertation 20 0.12 0.09 0.16 23.79 (19)  

Notes:aQw is based on the fixed-effect model.

bQB is based on the mixed-effect model.

cSchool level is based on 35 cases.

*Significant at α = 0.05; **significant at α = 0.001.

Table 4. 
Analyses of Semi-partial Correlations by Building Condition Characteristics
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Building condition measure      7.83 (2)* 
Cosmetic 0.11 0.03 0.19 0.09 (1)  
Structural 13 0.14 0.11 0.17 9.16 (12)  
Overall 21 0.07 0.03 0.11 48.51 (20)**  
Clusterc      1.99 (2) 
Temperature 0.11 0.05 0.18 0.16 (2)  
Physical 0.13 0.09 0.17 1.64 (5)  
Composite 26 0.09 0.05 0.13 73.59 (25)**  
Rater      0.93 (1) 
Internal personnel 21 0.11 0.07 0.15 31.38 (20)  
External personnel 15 0.08 0.04 0.13 44.83 (14)**  
Building instrument type      8.36 (1)* 
CAPE 16 0.14 0.11 0.16 8.45 (15)  
Other 20 0.06 0.02 0.10 59.53 (19)**  
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Building condition measure      7.83 (2)* 
Cosmetic 0.11 0.03 0.19 0.09 (1)  
Structural 13 0.14 0.11 0.17 9.16 (12)  
Overall 21 0.07 0.03 0.11 48.51 (20)**  
Clusterc      1.99 (2) 
Temperature 0.11 0.05 0.18 0.16 (2)  
Physical 0.13 0.09 0.17 1.64 (5)  
Composite 26 0.09 0.05 0.13 73.59 (25)**  
Rater      0.93 (1) 
Internal personnel 21 0.11 0.07 0.15 31.38 (20)  
External personnel 15 0.08 0.04 0.13 44.83 (14)**  
Building instrument type      8.36 (1)* 
CAPE 16 0.14 0.11 0.16 8.45 (15)  
Other 20 0.06 0.02 0.10 59.53 (19)**  

Notes:aQw is based on the fixed-effect model.

bQB is based on the mixed-effect model.

cCluster level is based on 35 cases.

*Significant at α = 0.05; **significant at α = 0.001.

Table 5. 
Analyses of Semi-partial Correlations by Methodological Characteristics
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Attendance      1.55 (1) 
Uncontrolled 26 0.11 0.07 0.15 42.78 (25)*  
Controlled 10 0.08 0.03 0.12 31.31 (9)**  
SES      18.20 (1)** 
Uncontrolled −0.16 −0.29 −0.04 0.93 (4)  
Controlled 31 0.11 0.08 0.13 67.26 (30)**  
95% CI for
CategoryNumber of EffectsLowerUpperQW(df)aQB(df)b
Attendance      1.55 (1) 
Uncontrolled 26 0.11 0.07 0.15 42.78 (25)*  
Controlled 10 0.08 0.03 0.12 31.31 (9)**  
SES      18.20 (1)** 
Uncontrolled −0.16 −0.29 −0.04 0.93 (4)  
Controlled 31 0.11 0.08 0.13 67.26 (30)**  

Notes:aQw is based on the fixed-effect model.

bQB is based on the mixed-effect model.

*Significant at α = 0.05; **significant at α = 0.001.

Academic Measure

First we examined whether or not the strength of the building condition–performance relationship varied as a function of the achievement outcome. Contrary to the results based on the bivariate correlation, results indicated that the type of academic measure was not related to the variability in effect sizes.

School Level

The test of the between-group differences by school level was statistically significant, indicating that the semi-partial correlations differed as a function of school level. Positive mean effects were seen in elementary school levels (0.11), whereas effects based on high school level revealed a negative mean effect (−0.21).

Publication Status

No significant variation was seen in group means for publication status. Under the fixed-effects model, however, effects based on published documents varied significantly.

Measure of Building Condition

The building condition category included cosmetic, structural, and overall condition measures. The groups were statistically different from zero (QB (2) = 7.83, p < 0.05), indicating that the measure representing building condition can explain some of the variability in the correlation. The mean ESs for structural and cosmetic features were higher compared with the mean effect of the overall/composite measure.

Rater Who Assessed Building Condition

The relationship of building condition to student academic achievement did not significantly differ as a function of the rater assessing the condition of buildings. Under the fixed-effects model, however, effects based on ratings from external personnel varied significantly.

Instrument Type

Again, our data included effect sizes from studies utilizing the CAPE or “other” instruments to assess the condition of school building. Statistically significant variations were seen between the two groups of instruments. The mean effect was higher in studies using the CAPE (0.14) compared with those using other instruments (0.06). In addition, significant variation was seen within the set of effects from other instruments.

Attendance

We were interested in identifying whether or not statistically controlling for student attendance within the primary studies impacted the semi-partial correlation. Contrary to what we expected, the test of between-group differences was not significant. Under the fixed-effects model, however, effects based on models both controlling and not controlling for attendance varied significantly.

Socioeconomic Status

We were also interested in identifying whether or not statistically controlling for SES within the primary studies impacted the semi-partial correlation. The semi-partial correlation mean significantly varied (QB (1) = 18.20, p < 0.001), as a function of whether or not regression models statistically controlled for student socioeconomic status. Models not controlling for SES as covariate had higher negative mean estimates (−0.16). In addition, the result was heterogeneous among the SES controlled group.

Study Year

Contrary to the results from correlational analyses, study year did not explain variation in the magnitude of the relationship under the mixed-effects model (QB (1) = 0.02, p > 0.05).

Publication Bias

The issue of publication bias has been well documented in the meta-analysis literature (Rosenthal 1979; Lipsey and Wilson 2001; Rothstein, Sutton, and Borenstein 2005). This issue occurs because statistically significant results are more likely to be published than nonsignificant results (Rothstein, Sutton, and Borenstein 2005). We evaluated both correlations and semi-partial indexes using three methods of detecting publication bias.

First, we conducted moderator analyses to determine the impact of publication status on the relationship between building condition and student performance. Correlations (see table 1), revealed statistically significantly stronger effects in published documents (0.21) than unpublished documents (0.10). The semi-partial correlations (see table 2) revealed significant differences between published and unpublished documents in the values. From this method, we can conclude that publication bias may exist in the correlation values.

Second, we constructed separate funnel plots of study sample size against the effect sizes, where the plus sign (+) and the circle (o) represent effect sizes from published and unpublished documents, respectively. Obtaining an asymmetrical or skewed shape indicates the potential presence of publication bias (Rothstein, Sutton, and Borenstein 2005). The plot of the correlations (see figure 3) showed an asymmetrical shape but the plot of the semi-partial correlations (figure 4) does not show a strong asymmetry in the values. In addition, Egger's linear regression test was statistically significant (t = 2.86, p < 0.05) for correlation data but not significant (t = 0.721, p > 0.05) for semi-partial values. Both results have confirmed that publication bias may exist in correlation studies.

Figure 3.

Funnel Plot of Correlations.

Figure 3.

Funnel Plot of Correlations.

Figure 4.

Funnel Plot of Semi-partial Correlations.

Figure 4.

Funnel Plot of Semi-partial Correlations.

5.  Discussion

For years, educational researchers have claimed that a relationship exists between building condition and student performance, but the extent of this relationship remained unclear. This review provides a better understanding of the degree of this relationship, using meta-analytic procedures.

The Overall Relationship between Building Condition and Student Achievement

The overall effect size of the correlation coefficient and the semi-partial correlation coefficients reflect a small level of association between building condition and academic achievement: and . The estimated means are small but significantly different from zero, implying that, on average, school building conditions are positively related to academic achievement to a weak degree. Although the average effects are small, the results should not be interpreted simply as improvement of the condition of the school buildings would have little to offer in enhancing student achievement. Instead, the results should be considered in conjunction with the breakdown analysis of study characteristics.

A possible explanation for the lack of a strong relationship may be explained by the moderating effects. Of importance, results represented by the bivariate correlation and the semi-partial correlation show the relationship varied as a function of the building feature measured. The relationship was stronger for cosmetic and structural components than composite scores including features of both cosmetic and structural features. These results coincide with the decades of research on multidimensional constructs. Research suggests that multidimensional constructs explain less variance than the variance explained by the dimensions (e.g., Humphreys 1962; Schmidt and Kaplan 1971; Rothstein, Sutton, and Borenstein 2005). The school facility is a multifaceted contrast—without establishing validity it is possible that scores based on the overall building conditions diluted the correlations.

The significant variation within the breakdown analysis of the building features measured based on the correlations suggests that certain features of the school facility may have more of an effect than other features on students’ academic achievement. In fact, the mean effects for noise level and features representing physical appearance based on the bivariate correlation coefficient was significantly larger than the mean effect from other features. Although building components as a whole may not impact student performance, it is clear that certain features such as noise can make it difficult for students to learn, think, and concentrate. Thus, including features that have a weak or nonsignificant relationship with student achievement may have caused the correlations to decline.

The conflicting results in terms of the instrument type and the rater assessing the building conditions likely emerged from the items assessed. Correlational studies reveal a larger effect using “other” instruments, while multiple regression analyses show correlations are larger when conditions are evaluated using the CAPE survey. The heterogeneity between studies raises concerns and supports the suggestion made by Earthman and Lemasters (2011) that further evaluations for reliability and validity are needed.

The Relationship in Terms of Student Characteristics

Beyond the main findings, the relationship varied as a function of the specific area of performance measures and grade levels. Correlations between building condition and academic performance increased and decreased in the measures of student performance. The direction and strength of the correlations changed depending on the grade level for results represented by the bivariate correlation and the semi-partial correlation.

The stronger and positive relation in younger students is not surprising because learning and development during primary school years are very complex (Wellman and Lagattuta 2004). In general, “elementary students lack the ability to deploy memory and thinking strategies effectively” (Cox et al. 2009, p. 2). Consequently, a noisy, uncomfortable, or nonorderly environment can fuel preexisting problems, making it more complicated for younger students to develop the intended skills, which in turn may lead to an impact on performance and a higher correlation.

From a psychometric perspective, another possible explanation for the moderating effects goes back to the validity of the measurement of conditions. Determining the influence of the school building conditions may require different measures for grade levels and specific areas of student performance. Although there have been a number of studies examining the relationship across different grade levels and student outcomes, few studies have examined the effect of such variables on the predictive validity of conditions. Thus, future research is needed to determine the validity by which the conditions are related to specific achievement measures and grade levels.

The Relationship in Terms of Controlling for Influential Variables

Eighty-seven percent of the semi-partial correlations included in our analyses were computed from models that controlled for SES, and 27 percent from models controlling for student attendance. Our findings show that including SES into the model dilutes the effect of the school building conditions, and controlling for student attendance appears to have little impact on the relationship.

The fact that including SES diluted the effect for the school building conditions is not surprising given the causal nature of SES with student achievement (Sirin 2005). Although controlling for student attendance appeared to have little impact, it is possible there are influential factors beyond student attendance that influence the relationship. For example, a number of studies suggest parental involvement is a vital component for enhancing student achievement (Epstein 2001; Jeynes 2005; Houtenville and Conway 2008). Thus future research needs to statistically control for variables influencing student performance.

Taken together, the results from this review are very important given the thin knowledge in this area. The results support the claim made in prior reviews that there does exist a small positive relationship between building condition and student performance. Although our findings suggest that there is a weak relationship between building condition and student performance, the review highlights important facets affecting the strength of the relationship. The broad range of statistical significant effect sizes suggests the need to focus on the structural validly of the assessment tools. Given that the school facility is a multifaceted construct, the ability to measure and assess the quality of the building conditions requires the establishment of structural validly. Without establishing structural validity, it is difficult to reflect the influence of the school building conditions on student performance.

6.  Strengths and Limitations

The major strength of this review is that it provides a better picture of the overall relationship between building condition and academic achievement. Nonetheless, our review includes some weaknesses and the results should be interpreted with caution. First, the studies included in the meta-analysis were nonexperimental designs and most likely suffered from selection bias.

Second, because our data set included multiple effects from a single study, we violated the independent assumption underlying meta-analysis. However, because we examined moderator factors to identify the impact of study characteristics on this relationship, we reduced type I error.

Third, from the empirical test of the moderator variable publication status, the inspection of the funnel plot, and Egger's regression test, publication bias may exist in the r values. Results from multiple regression analyses revealed no significant difference between published and unpublished documents. Although it is possible that publication bias exists, another possible explanation relates to the fact that the majority of studies included in the review were from dissertations. It is possible that there is a lack of graduates submitting manuscripts for publication. Thus, we exhort professors to encourage students to publish their scholarly work.

Fourth, it is possible that this meta-analysis does not include all studies examining this relationship. Although we searched many electronic databases and examined cited narrative reviews to locate relevant studies, some studies could not be obtained. Therefore, our sample may not represent all populations.

Finally, meta-analyses are limited by the quality of the research included in the synthesis (Glass, McGaw, and Smith 1981). Many studies in this review did not report the validity or reliability of instruments used to evaluate the condition of school facilities. Additionally, many regression models did not control for other important variables such as teacher quality or parental involvement. The quality of instruments and designs of primary studies can be detrimental, thus results should be interpreted with caution.

7.  Implications for Future Research

Despite the limitations, the results from this synthesis can be a guide for researchers. The lack of a strong relationship suggests additional research and development is required for a more complete assessment of the condition of the school buildings. On the basis of our findings, obtaining a composite score from multiple components may underestimate the effect of building condition. Because different components of building conditions yield different results, researchers should not obtain a composite score of cosmetic and structural features. Instead, researchers should look at each aspect separately. This in turn will provide educators and policy makers a deeper level of understanding of the features most essential for enhancing student learning.

Second, the findings of this review show that results vary between and within instruments designed to assess the condition of school facilities. The heterogeneity could be partly explained by the presence of items that are not related to student achievement. The inconsistency raises concerns about the structural validity of the assessment tools. Therefore, future research is needed to develop a universal instrument that captures features of school facilities that predict student performance or establish reliability and validity for instruments being used.

Third, many of the primary studies included in this review did not control for variables that have been shown to influence student performance. For instance, research suggests there is a strong relationship between SES and student achievement (Sirin 2005), and parental involvement and student achievement (Fan and Chen 2001; Jeynes 2005). When examining these relationships using more complex analyses such as multiple regressions, we recommended researchers control for such variables. Not controlling for variables that influence student performance can impact the conclusions which primary researchers and meta-analysts derive.

Finally, meta-analysis has become increasingly popular in many areas, with researchers proposing many indices to synthesize more complex analyses (Keef and Roberts 2004; Becker and Wu 2007; Kim 2011; Aloe and Becker 2012). In order to support the work of future meta-analysts, we highly recommend researchers report sufficient information to compute effect sizes, such as the correlation matrix, standardized regression coefficients and their standard errors, sample sizes, and the multiple correlation coefficients.

8.  Implications for Educational Leaders and Policy Makers

As policy makers continue to make critical decisions to improve student performance, the impact of the condition of the school buildings on student performance is of interest to many educational leaders and policy makers. Over 50 percent of K–12 public schools are in need of repairs, renovations, and modernizations. According to the U.S. Department of Education, approximately $197 billion is needed to improve the condition of the school buildings (USDOE 2014).

Measuring and reporting on the influence of the school building conditions on student performance is important; it aids educational leaders and policy makers in decision making and resource allocations. From a practical perspective, the lack of a strong relationship may suggest improving the condition of the school buildings would have little impact on student performance. Practitioners should be cautious in interpreting the results, however, given the various moderator effects. Issues relating to items designed to measure the school building condition can cause the relationship to decline, impacting the evidence on the importance of the building conditions for enhancing student performance.

9.  Conclusion

For many years researchers have claimed a relationship exists between building condition and student performance. The results in this review indicate that a small but overall positive relationship does exist. Furthermore, the relationship varied as a function of various contextual and methodological characteristics.

The findings of this synthesis provide a better insight into the relationship between building and student performance and a thorough examination of factors impacting the relationship. This synthesis offers useful information for educational leaders, policy makers, and researchers interested in this area.

Note

*

Studies included in this meta-analysis

Acknowledgments

The authors would like to thank Dr. Betsy J. Becker for all her support and assistance.

REFERENCES

*
Al-Enezi
,
M.
2002
.
A study of the relationship between school building conditions and academic achievement of twelfth grade students in Kuwaiti public high schools
.
PhD dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA
.
Aloe
,
Ariel M.
, and
Betsy J.
Becker
.
2012
.
An effect size for regression predictors in meta-analysis
.
Journal of Educational and Behavioral Statistics
37
(
2
):
278
297
.
Bailey
,
J.
2009
.
A synthesis of studies pertaining to school building condition, student achievement, student behavior, and student attitude
.
PhD dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA
.
Becker
,
Betsy J.
, and
Meng-Jia
Wu
.
2007
.
The synthesis of regression slopes in meta-analysis
.
Statistical Science
22
(
3
):
414
429
.
*
Berner
,
Maureen M.
1993
.
Building conditions, parental involvement, and student achievement in the District of Columbia public school system
.
Journal of Urban Education
28
(
1
):
6
29
.
Cash
,
C.
1993
.
Building condition and student achievement and behavior
.
PhD dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA
.
*
Cervantes
,
R.
1999
.
The condition of school facilities as related to student academic achievement and behavior
.
PhD dissertation, University of Alabama at Birmingham, Birmingham, AL
.
Cox
,
Brian
,
David
Geary
,
Ray
Brogan
,
Dorothy
Singer
, and
Mary
Gauvain
.
2009
.
Cognitive development
. Available www.education.com/reference/article/cognitive-development2.
Accessed 10 May 2014
.
*
Duran-Nurucki
,
Valkiria
.
2008
.
School building condition, school attendance, and academic achievement in New York public schools: A mediation model
.
Journal of Environmental Psychology
28
(
3
):
278
286
.
Earthman
,
Glen
.
2004
.
Prioritization of 31 criteria for school building adequacy
. [Executive Summary]. Available www.schoolfunding.info/policy/facilities/ACLUfacilities_report1-04.pdf.
Accessed 29 August 2014
.
Earthman
,
Glen I.
, and
Linda K.
Lemasters
.
2011
.
The influence of school building conditions on students and teachers: A theory-based research program
(
1993–2011
).
The ACEF Journal
1
(
1
):
15
36
.
Epstein
,
Joyce
.
2001
.
School, family, and community partnerships: Preparing educators and improving schools.
Boulder, CO
:
Westview Press
.
*
Evans
,
Gary
,
John
Sipple
, and
Min Jun
Yoo
.
2010
.
The ecological context of student achievement: School building quality effects are exacerbated by high levels of student mobility
.
Journal of Environmental Psychology
30
(
2
):
239
244
.
Fan
,
Xitao
, and
Michael
Chen
.
2001
.
Parental involvement and students’ academic achievement: A meta-analysis
.
Educational Psychology Review
13
(
1
):
1
22
.
Ferillo
,
Bud
.
2006
.
Corridor of shame. The neglect of South Carolina's rural schools
(film). Available www.corridorofshame.com./
Accessed 17 February 2015
.
Fisher
,
R. A.
1924
.
On a distribution yielding the error functions of several well known statistics
.
Proceedings of the International Congress of Mathematics, Toronto
2
:
805
813
.
Glass
,
Glass V.
,
Barry
McGaw
, and
Mary L.
Smith
.
1981
.
Meta-analysis in social research.
Beverly Hills, CA
:
Sage
.
Goldhaber
,
Dan
, and
Dominic
Brewer
.
1997
.
Why don't schools and teachers seem to matter? Assessing the impact of unobservables on educational productivity
.
Journal of Human Resources
32
(
3
):
505
523
.
Guy
,
L.
2001
.
Student achievement and school condition: Examining the relationship in West Virginia's high schools
.
PhD dissertation, West Virginia University, Morgantown, WV
.
Hedges
,
Larry V.
, and
Ingram
Olkin
.
1985
.
Statistical methods for meta-analysis
.
Orlando, FL
:
Academic Press
.
Hedges
,
Larry V.
, and
Therese
Pigott
.
2004
.
The power of statistical tests for moderators in meta-analysis
.
Psychological Methods
9
(
4
):
426
445
.
Hines
,
E.
1996
.
Building condition and student achievement and behavior
.
PhD dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA
.
Houtenville
,
Andrew J.
, and
Karen
Conway
.
2008
.
Parental effort, school resources, and student achievement
.
The Journal of Human Resources
43
:
437
453
.
Humphreys
,
Lloyd G.
1962
.
The organization of human abilities
.
American Psychologist
17
(
7
):
475
483
.
Jeynes
,
William H.
2005
.
A meta-analysis of the relation of parental involvement to urban elementary school student academic achievement
.
Urban Education
40
(
3
):
237
269
.
Kahneman
,
Daniel
, and
Amos
Tversky
.
1984
.
Choices, values and frames
.
American Psychologist
39
:
341
350
.
Keef
,
Stephen P.
, and
Leigh
Roberts
.
2004
.
The meta-analysis of partial effect sizes
.
British Journal of Mathematical and Statistical Psychology
57
(
1
):
97
129
.
Kim
,
R.
2011
.
Standardized regression coefficients as indices of effect sizes in meta-analysis.
PhD dissertation, Florida State University, Tallahassee, FL
.
Lamdin
,
Douglas J.
1996
.
Evidence of student attendance as an independent variable in education production functions
.
The Journal of Educational Research
89
(
3
):
155
162
.
Lemasters
,
Linda K.
1997
.
A synthesis of studies pertaining to facilities, student achievement, and student behavior
.
PhD dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA
.
Lipsey
,
Mark W.
, and
David B
Wilson
.
2001
.
Practical meta-analysis
.
Thousand Oaks, CA
:
Sage Publications
.
McGuffey
,
Carroll
.
1982
.
Facilities
. In
Improving educational standards and productivity: The research basis for policy
, edited by
Herbert J.
Walberg
, pp.
237
288
.
Berkeley, CA
:
McCutchan Publishing
.
*
O’Sullivan
,
Sean
.
2006
.
A study of relationship between building conditions and student academic achievement in Pennsylvania's high school
.
PhD dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA
.
*
Picus
,
Lawrence O.
,
Naomi
Calvo
,
William J.
Glenn
, and
Scott
Marion
.
2005
.
Understanding the relationship between student achievement and the quality of educational facilities: Evidence from Wyoming
.
Peabody Journal of Education
80
(
3
):
71
95
.
Rosenthal
,
Robert
.
1979
.
The file drawer problem and tolerance for null results
.
Psychological Bulletin
86
(
3
):
638
641
.
Rothstein
,
Hannah R.
,
Alexander J.
Sutton
, and
Michael
Borenstein
.
2005
.
Publication bias in meta-analysis
. In
Publication bias in meta-analysis: Prevention, assessment, and adjustments
, edited by
Hannah R.
Rothstein
,
Alexander J.
Sutton
, and
Michael
Borenstein
, pp.
1
7
.
Chichester, UK
:
John Wiley & Sons, Ltd
.
Schneider
,
Mark
.
2003
.
Linking school facility conditions to teacher satisfaction and success.
Available http://eric.ed.gov/?id=ED480552.
Accessed 3 August 2010
.
Schmidt
,
Frank L.
, and
Leon B.
Kaplan
.
1971
.
Composite vs. multiple criteria: A review and resolution of the controversy
.
Personnel Psychology
24
(
3
):
419
434
.
Sirin
,
Selcuk R.
2005
.
Socioeconomic status and academic achievement: A meta-analytic review of research
.
Review of Educational Research
75
(
3
):
417
453
.
Stewart
,
Ron L.
2010
.
A meta-analytical synthesis of studies on the effects that building conditions, building age, artificial lighting, and natural lighting in schools has on student learning and behavior
.
PhD dissertation, George Fox University, Newberg, OR
.
U.S. Department of Education (USDOE)
.
2014
.
Condition of America's public school facilities: 2012–13
. Available http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2014022.
Accessed 5 June 2014
.
Weinstein
,
Carol S.
1979
.
The physical environment of the school: A review of the research
.
Review of Educational Research
49
(
4
):
577
610
.
Wellman
,
Henry M.
, and
Kristin H.
Lagattuta
.
2004
.
Theory of mind for learning and teaching: The nature and role of explanation
.
Cognitive Development
19
(
4
):
479
497
.