Full
Spectrum Lighting
Full Spectrum Lighting
|
|
Daylighting in Schools An Investigation into the Relationship Between Daylighting and Human Performance Condensed Report August 20, 1999 Submitted to: George Loisos The Pacific Gas and Electric Company on behalf of the California Board for Energy Efficiency Third Party Program Submitted
by: HESCHONG
MAHONE GROUP Legal Notice This report was prepared by Pacific Gas and Electric Company and funded by California utility customers under the auspices of the California Public Utilities Commission. Neither
PG&E nor any of its employees and agents: (2) assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, process, method, or policy contained herein; or (3) represents that use of the report would not infringe any privately owned rights, including, but not limited to, patents, trademarks, or copyrights. © 1999 by Pacific Gas and Electric Company. All rights reserved. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP i August 20, 1999 ACKNOWLEDGEMENTS This study was performed on behalf of the California Board for Energy Efficiency for the Third Party Program administered by Pacific Gas and Electric, as part of PG&E contract 460 000 8215. George Loisos was the project manager and Mona Yew the Contract Administrator. Lisa Heschong, Partner in the Heschong Mahone Group, directed the study. She has been assisted at the Heschong Mahone Group by Douglas Mahone, Kalpana Kuttaiah, Nehemiah Stone, Cathy Chappell, Jon McHugh, and Jackie Burton. Stacia Okura of RLW Analytics conducted the statistical analysis under the direction of Dr. Roger Wright, Principal, RLW Analytics. Barbara Erwine of Cascadia Conservation investigated daylighting conditions at the Seattle Public Schools. Neal Digert and Ken Baker of Architectural Energy Corporation investigated daylighting conditions at Poudre School District in Fort Collins, Colorado. Both Cascadia Conservation and Architectural Energy Corporation participated in data acquisition and development of the analysis methodology. We are deeply indebted to the school district personnel who made this study possible, by providing data and allowing us access to district records and facilities. Jeff Bristow and Chuck Berridge at Capistrano Unified School District in Southern California, Mike O’Connell at Seattle City Public Schools in Washington State and Hugh Mowery at Poudre School District in Fort Collins, Colorado, provided access to their district’s data and assisted with its interpretation. Dave Doomey, Ken Harkner and Bob Sendzik at Capistrano, Kathy Johnson at Seattle and Mike Spearnak at Poudre helped provide information about and access to their district facilities. We are very thankful to the many other people who also made this study possible, through their interest in the significance of this work and their willingness to provide helpful information and support. We would especially like to thank those who took the time to review and comment on the draft reports: Gregg Ander, Dr. Ed Arens, Dr. Gale Berger, Dr. Robert Clear, Dr. Rick Diamond, Dr. Judith Heerwagen, Dr. Paul Holland, Dr. Gage Kingsbury, Eleanor Lee, Dr. Margaret Morris, and Dr. David Wyon; and Steven Selkowitz who organized the review. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 3 August 20, 1999 EXECUTIVE SUMMARY This study looks at the effect of daylighting on human performance. It includes a focus on skylighting as a way to isolate illumination effects from other qualities associated with daylighting from windows, such as view and ventilation. In this project, we established a statistically compelling connection between daylighting and student performance, and between skylighting and retail sales. This report focuses on the school analysis. We obtained student performance data from three elementary school districts and looked for a correlation to the amount of daylight provided by each student’s classroom environment. We used data from second through fifth grade students in elementary schools for two reasons: because there is extensive data available from highly standardized tests administered to these students, and because elementary school students are generally assigned to one teacher in one classroom for the school year. Thus, we reasoned that if the physical environment does indeed have an effect on student performance, we would be mostly likely to be able to establish such a correlation by looking at the performance of elementary school students. We analyzed test score results for over 21,000 students from the three districts, located in Orange County, California, Seattle, Washington, and Fort Collins, Colorado. The data sets included information about student demographic characteristics and participation in special school programs. We reviewed architectural plans, aerial photographs and maintenance records and visited a sample of the schools in each district to classify the daylighting conditions in over 2000 classrooms. Each classroom was assigned a series of codes on a simple 0-5 scale indicating the size and tint of its windows, the presence and type of any skylighting, and the overall amount of daylight expected. The study used multivariate linear regression analysis to control for other influences on student performance. Regressions were compared using data from two separate tests, math and reading, for each district. Each math and reading model was also run separately using first the window and skylight codes, and then the overall daylight code. We reasoned that if daylight effects were truly robust the variables should perform similarly in all models. Thus, we created a total of twelve models for comparison, consisting of four models for each of three districts. The daylighting conditions at the Capistrano school district were the most diverse, and the data from that district were also the most detailed. Thus Capistrano became our most precise model. In this district, we were able to study the change in student test scores over a school year. Controlling for all other influences, we found that students with the most daylighting in their classrooms progressed 20% faster on math tests and 26% on reading tests in one year than those with the least. Similarly, students in classrooms with the largest window areas were found to progress 15% faster in math and 23% faster in reading than those with the least. And students that had a well-designed skylight in their room, one that diffused the daylight throughout the room and which allowed teachers to c than those students without a skylight. We also identified another window-related effect, in that students in classrooms where windows could be opened were found to progress 7-8% faster than those in rooms with fixed windows. This occurred regardless of whether the classroom also had air conditioning. These effects were all observed with 99% statistical certainty. The studies in Seattle and Fort Collins used the final scores on math and reading tests at the end of the school year, rather than the amount of change from the beginning of the year. In both of these districts we also found positive, and highly significant, effects for daylighting. Students in classrooms with the most daylighting were found to have 7% to 18% higher scores than those in rooms with the least. The three districts have different curricula and teaching styles, different school building designs and very different climates. Yet the results of the studies show consistently positive and highly significant effects. This consistency supports the proposition that there is a valid and predictable effect of daylighting on student performance. The results of this study of student performance, when considered along with those of the companion study showing the positive effect of skylighting on retail sales, also strongly support the thesis that these performance benefits from daylighting can be translated to other building types and human activities. 1. DAYLIGHTING IN SCHOOLS This report is part of a study that looks at the effect of daylighting on human performance. This part of the study looks at how daylighting, from windows or skylights, affects the test scores of students in three elementary school districts. Another part of the study reports on how the use of skylighting affected the sales of a large chain retailer. We found a statistically compelling connection between daylighting and student performance, and between skylighting and retail sales. The study was supported by the California Board for Energy Efficiency, and administered by Pacific Gas and Electric Company. Schools and retail stores were chosen as the subject of the study because we could obtain extensive data on occupant performance for nearly identical buildings. We believe that the conclusions may be transferable to other types of buildings, such as offices and factories, since it is really human performance that we are investigating. If daylighting enhances the performance of children in schools, it is not too large a stretch to suppose that it might also enhance the performance of adults in office buildings. If daylighting motivates buyers at a retail store, it is not too large a stretch to presume that it might also motivate workers in a factory. This Condensed Report is intended for the non-specialist reader. It is a summary of a more extensive report that details the study methodology and statistical analysis. If you have questions about the study that are not answered here, we recommend reading the Detailed Report. 1.1 Background The impact of daylighting on the performance of school children has been a subject of interest for many years. Before fluorescent lighting became prevalent, it was generally assumed that all school rooms would be daylit as a matter of course. The California Department of Education had a rigorous review process for the architectural design of classrooms to ensure that daylighting standards were met. As a result, California classrooms built in the 1950’s and early 1960’s remain excellent examples of daylighting practice. The "finger" plan with multiple rows of single classrooms, each with windows on two sides, became a standard for California K-12 campuses. However, starting in the late 1960’s a number of forces came into conflict with the daylit design of classrooms. Engineers, asked to provide air conditioning in classrooms, argued against the use of large expanses of glass and high ceilings. Construction economists argued that schools could be built more inexpensively on smaller sites if the classrooms could be built back to back or grouped together, without constraints on solar orientation. Educational theorists argued that a more flexible arrangement of classrooms, with open walls between them, would encourage team teaching and creative learning. And educational planners, trying to meet the needs of an exploding school age population, required that at least one-third of all new classrooms be portable, so that, if the need arose, they could be moved to new areas with an overpopulation of new students. As a result of these various pressures, the finger plan school was largely abandoned in California, and a vast experimentation in school design was undertaken. Many of the classrooms built since the 1960’s have little daylighting. Windows are commonly built with "black glass" that allows a view out, but no useful daylight in. Numerous schools have been built with no windows at all. Similar trends occurred nationally, and internationally, though perhaps without such a dramatic shift in design practice as in California. Concerned about the trend towards schools, and all types of buildings, without windows, Belinda Collins of the National Bureau of Standards conducted a major literature review on the study of windows in 19741. At that time there was an ongoing debate about the desirability of windows in classrooms. In a compilation of studies on windowless classrooms published in 1965, the editor, C.T. Larson, concluded that windowless classrooms should have no adverse effects upon their users. Larson stated, "The educational value of such a view [that windows are necessary for student learning] should be assessed against the cost of installing and maintaining classroom windows.2" Collins also quotes from a later book on the behavioral aspects of design, which also concluded that windows were not needed in classrooms. "At present the pro-window forces still lack behavioral data in support of their case and argue on the basis of metaphor and supposition, but their arguments must be weighed against statistics…from the windowless schools…reported to have 40 percent greater efficiency in heating and cooling, constant light to prevent eye strain…35 decibels or more noise reduction, and reduced maintenance costs." The author went on to claim that the experience of completely underground schools provided evidence that claustrophobic reactions were extremely rare. He stated further that, "Opponents [of windowless schools] now take recourse in the need for communion with nature, contact with the outside and stimulus variation, which are more difficult to measure, and whose importance is not readily apparent." Collins herself found that the research that had been done as of 1974 was suggestive of the importance of windows, but inconclusive: "Much, though not all, of the evidence from the windowless classroom studies is inconclusive, or inadequate, while that from windowless factories is circumstantial, based on hearsay, rather than research. As a result, only tentative conclusions can be drawn about the qualities of windowless spaces that make them somewhat less than desirable." 1 Collins, B. "Windows and People: a Literature Survey, Psychological Reaction to Environments With and Without Windows", National Bureau of Standards, June 1975 2 Larson, C.T. (ed), The Effect of Windowless Classrooms on Elementary School Children, The Architectural Research Laboratory, Department of Architecture, University of Michigan, 1965. Since Collins’ study, other research on the importance of windows has been done, but primarily in hospitals. The most rigorous studies have been conducted in Europe. One interesting study in Sweden in 1992 looked at the impact of daylight on the behavior of elementary school children. The Swedish researchers followed the health, behavior, and hormone levels of 88 eight-year-old students in four classrooms over the course of one year. The four classrooms had very different daylight and electric light conditions: two had daylight, two had none; two had warm white (3000K) fluorescent lamps, two had very cool (5500K) fluorescent lamps. The researchers found significant correlation between daylight levels, hormone levels, and student behavior, and concluded that windowless classrooms should be avoided. Recent, more informal studies in the United States claiming a relationship between daylighting and enhanced student performance have generated considerable excitement among daylighting advocates. These studies, along with a rising interest in "natural" and "healthy" environments, have contributed to a resurgent interest in daylighting in schools. All three districts that we worked with in this study reported that daylighting in classrooms is currently a concern for their school boards, driven largely by parent activism. However, without credible evidence of relationship between the design of schools and the performance of students within them, classroom design issues remain subject to architectural and educational fads, just as in the past. We hope that this study provides a contribution towards more durable understanding of how the physical environment affects student performance. 1.2 The School Data We obtained student performance data from three elementary school districts and looked for a correlation between test scores and the amount of daylight provided by each student’s classroom environment. We used data from second through fifth grade students in elementary schools because there is extensive data available from highly standardized tests administered to these students, and because elementary school students are generally assigned to one teacher in one classroom for the school year. Thus, we reasoned that if the physical environment does indeed have an effect on student performance, it would be most apparent in populations of elementary school students. We analyzed test score results for over 21,000 students from the three districts, located in Orange Country, California; Seattle, Washington; and Fort Collins, Colorado. The three districts have different curricula, administrative and teaching styles, different school building designs and very different climates. Yet the results of the studies show consistently positive and highly significant effects. This consistency supports the proposition that there is a valid and predictable effect of daylighting on student test scores. The districts provided us with a wide variety of data sets, with many different student test scores and student demographic characteristics, for a two year period. In order to achieve consistency between districts, we chose to use the data from just two test scores, reading and math, in our analysis. We also endeavored to keep the demographic variables consistent between districts. Individual student identities were masked by substituting false student record numbers for all data sets. In addition, some districts decided to provide some demographic data at a classroom level to further mask individual student records. Similarly, in our reporting, we have scrambled the identification numbers for school sites, and renamed the specific demographic variables in this report to make them generic. A second data set was created describing the physical characteristics for each classroom in the three districts. This data allowed us to take into account the age and size of the classroom and school, the type of the classroom, (open, cluster or pod, portable or traditional) as well as the presence and size of windows and skylights. We reviewed architectural plans, aerial photographs and maintenance records and visited a sample of the schools in each district to classify the daylighting conditions in over 2000 classrooms. Each classroom was assigned three codes on a simple scale indicating the size and tint of its windows, the presence and type of any skylighting, and the overall amount of daylight provided. In this section we report on the findings for each of the three school districts in turn. First we describe the relevant characteristics of each district, so that the reader can understand the context and better evaluate the results. Then we report on the specific model results for each district. The study used a powerful statistical analysis technique, called multivariate linear regression analysis, to control for other influences on student performance. These mathematical models allow us to isolate the effect of one variable, while controlling for the influence of all the others. The models also tell us the statistical probability that we have a "true" effect, and the power of each variable in predicting results. With the Capistrano data, we created a model based on the change in test scores between the fall of 1997 and spring of 1998. Thus, this analysis looks at the rate of learning during the school year that the students occupied a given physical environment. It also uses each student as his or her own control. As a result, all of the demographic variables drop out, and we are left with a simple model containing only those few variables that are seen to directly influence student improvement. For the other two districts we had to use only the final spring test scores, rather than the difference between a fall and spring test. The models for these two districts report on a snapshot of student performance at one point in time. There is an assumption that the most recent classroom experience will influence how students perform on tests. However, the absolute level of student performance is a function of many influences, including where each student started at the beginning of the year and all the advantages or disadvantages that the students brought with them into the classroom. Thus, in these models, the demographic and socio-economic variables become important predictors of absolute student performance, and add many more control variables to our final equation. The Capistrano District provided by far the most complete and complex data set that we analyzed. We had the most information about its diversity in student population, administrative structure, and physical conditions. In the Capistrano analysis we were also able to account for the influence of the individual school, and to test for the influence of the individual classroom environment. Thus, we have the highest confidence in the results from the Capistrano study. The Seattle and Fort Collins studies are very suggestive of a daylighting effect on performance, but are not as exhaustive in their analysis or conclusive in their findings. It is the consistency of the positive findings from all three districts that makes a strong case that we have, indeed, found a valid effect. We also collected and analyzed information about the presence of air conditioning and operable windows in the classroom. We would have liked to include information about the different types of electrical lighting used in the schools, but this information was not available. Capistrano schools use fluorescent lighting throughout the district, and lighting systems are generally designed to provide an average of 50 footcandles of light on classroom work surfaces. However, there have been so many remodels and retrofits of the electric lighting system in recent years that the actual equipment type is highly variable. 2.1.2 Capistrano Results Figure 3 summarizes the increases in test scores for the daylighting-related variables for the four Capistrano regression models. As part of the analysis we calculated the statistical certainty that these effects were a "true" effect which could be replicated in other analyses of the data. This is expressed as a percent certainty. The chart shows the value of each variable’s effect, its statistical certainty, and the relative effect of each variable compared to the average progress of all students in the Capistrano District. Capistrano NEA Core Level Tests Range: -29 to +79 Change, Fall to Spring Reading Math Reading Math Reading Math Model 1 Daylight, Min. to Max. 2.8 2.3 99.9 99.9 26% 20% Operable Windows 0.8 - 99.8 n/s 7% - Model 2 Windows, Min. to Max. 2.4 1.7 99.9 99.9 23% 15% Skylight A 2.0 2.3 99.7 99.9 19% 20% Skylight B -2.2 - 94.9 n/s -21% - Operable Windows 0.9 0.8 99.6 99.9 8% 7% Difference in Average Test Improvement (normalized RIT points) Statistical Certainty Difference as a % of District Average Improvement Percentage Effect Analysis Results Figure 3: Summary Daylight Findings for Capistrano The Capistrano Core Level Tests are reported on a special scale system called "RIT." The average student in our data set progressed in reading scores by 8.8 RIT points and in math scores by 12.5 points from fall to spring5. For the charts in this report we have translated all the test results into a consistent scale of 1-99 in 5 These values are averages for our specific data set, not the district, because our data set was a sub-set of all students in the district. For the percentage effects discussed here, the raw RIT score (not the normalized score shown in the chart) was divided by this average from our data set. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 13 August 20, 1999 order to facilitate comparison between the districts. We also report the test results as a percentage effect to show the relative magnitude of the findings. Daylighting was found to have a considerable effect in the Capistrano schools. For example, all other things being equal, students in classrooms with Skylight Type A were found to progress an additional 2 points in reading and 2.3 points in math6 than those in classrooms without skylights. This translates into a 19% faster learning rate for reading and a 20% faster learning rate for math on average for the children in those classrooms. Summary results in the Capistrano Unified School District: w The classrooms with the most amount of daylighting are seen to be associated with a 20% to 26% faster learning rate, as evidenced by increased student test scores over one school year, compared to classrooms with the least amount of daylighting. w The classrooms with the most window area are seen to be associated with 15% to 23% faster rate of improvement over a one year period when compared to classrooms with the least amount of windows. w The classrooms with the Skylight Type A are seen to be associated with a 19% to 20% faster improvement when compared to classrooms with no skylights. w The classrooms with the Skylight Type B are seen to be associated with a 21% decrease for reading tests, and no significant results for math tests, when compared to classrooms with no skylights. w Classrooms with operable windows are seen to be associated with 7% to 8% faster improvement in three out of four cases, when compared to classrooms with fixed windows. Another way to look at these results is that the average child in the Capistrano district is making about 1 point of progress per month on the reading test and 1.5 points of progress per month on the math test over the course of the approximately eight months between the fall and the spring tests. Students in the most daylit classrooms are progressing more quickly, gaining one to two points more over the course of the school year than students advancing at the average rate. Thus, by advancing more quickly, students in daylit classrooms could save up to one month of instruction time in the reading and math curriculum that could be used for other areas of learning. 2.1.3 Other Capistrano Variables The results for all major variables of the Capistrano models are presented below in Figure 4. For this chart the values of the analysis results (called the Bcoefficient) have been normalized to a scale of 1-99 so that they can be 6 These are the normalized RIT values. Raw RIT values are 1.7 and 2.6 respectively. Thus, a 1.7 difference in reading scores, divided by the 8.8 district average, equals a 19% effect. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 14 August 20, 1999 compared with the other two school districts. The same set of control variables was considered in all regression models, and thus when a control variable was significant in all four Capistrano models, it has four bars in the graph. The Daylight, Window and Skylight variables each were run in only two of the four models, and thus, they have a maximum of two bars. We discuss the patterns and magnitudes of these findings below. Capistrano -4.00 -2.00 0.00 2.00 4.00 6.00 8.00 10.00 12.00 Daylight 0-5 Window 0-5 Skylight A Skylight B Oper. Window GATE 2nd Grade 3rd Grade 4th Grade Language Prog 500+ students 10 Absences Highest School Lowest School B Coefficient - Normalized RIT Difference from Change in Mean Score Reading Daylight Reading Skylight Math Daylight Math Skylight Figure 4: Capistrano Relative Importance of Variables, Normalized RIT Points, Difference From Change in Mean Score Each value predicts how far a given student is likely to diverge from the norm if the variable applies. It is very important to keep in mind that the Capistrano models use the rate of change in test scores over a school year as their measure, not the absolute levels of testing. Thus, a negative value for a variable means that those students made slightly less progress than the norm, but they still made progress. Daylight, Skylights, and Windows: The daylight, window and skylight type A variables are all positive and highly significant. Skylight Type A had the most even light distribution of the five skylight types, fully diffused without any potential for direct sunlight to enter the room. It also allowed CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 15 August 20, 1999 the teacher to control the amount of daylight with the use of manually controlled louvers. The observation that both the daylight variable and the Skylight Type A variable have slightly larger effects than the window variables argues for the theory that the presence of daylight in and of itself, and not view or other aspects of windows, are responsible for the positive effects. The results for the other skylight types were less compelling. The negative effect for Skylight Type B that is observed in one model might reasonably be interpreted as a function of the glare caused by sunlight splashing on the classroom walls. Skylight Type B is a clear acrylic skylight located in the corner of the classroom, often over the teacher’s desk. It is not provided with any controls to modulate the light. Thus, on sunny days, sunlight makes its way directly onto the walls or the teacher’s desk. This finding suggests that control of light and/or diffusion of direct sunlight are important features to include in a classroom skylight system. The other three skylight types, AA, C and D, had no significant coefficients. These skylights do not diffuse the light as evenly as skylight A, and in fact these skylights were often closed by the teachers. Thus, from these findings, it would seem that the mere presence of a "patch of daylight" or "connection to the outdoors" through toplighting is not sufficient to provide positive effects. The one skylight (type A) that performs well provides high levels of illumination, evenly distributed in the classroom. It does not allow any direct sunlight into the classroom, and also allows the teacher to easily modulate the light levels. Operable windows were also found to have a significant positive coefficient for three out of four of the models. We posit that allowing the teacher the option of using natural ventilation when desired is a positive feature for classrooms. About half of the classrooms with operable windows in this district also had air conditioning. However, in some of our preliminary analysis air conditioning in this district seemed to be associated with a negative effect. There are many possible interpretations of these findings, including interactions with other variables, the mild climate in Capistrano, malfunctioning air conditioning units, or air quality issues. We would suggest that this finding deserves further study. Grade Level: The grade level of the student tended to be the most powerful predictor of progress made between the fall and the spring tests. This is consistent with the RIT scales of the Capistrano core level tests, where younger grades typically make greater progress. In addition, California has recently mandated class size reduction for kindergarten through third grades so that students in the lower grades can receive more attention from their teachers. The maximum student/teacher ratio in those grades is 20:1, whereas in the higher grades in our data set, fourth and fifth, the ratio is commonly 30:1. Gifted and Talented (GATE) and Bilingual Programs: Participation in a GATE program shows a negative effect, meaning that GATE identified children made CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 16 August 20, 1999 slightly less progress in a year than non-GATE children. The best explanation of this would seem to be that GATE children already score very high on the tests. Since they are already at the top of the group, it is more likely, given all the variation in the system, that their scores will rise less quickly than others. This is also consistent with the observation that, in the RIT scaled tests, children testing at higher levels make less relative progress per year than those at lower levels. The positive effect of the bilingual program might be attributable to two further explanations, other than the obvious conclusion that the program is helping children progress more rapidly. Since the bilingual program children tend to have slightly lower actual scores than the norm, they would tend to progress faster than the norm. Alternatively, since the bilingual programs are magnet programs, they may attract more dedicated families, creating a self-selection bias for this population. School Site: Approximately 1/3 to ½ of the schools showed up in the models as having a significant influence on how much a student learned over the course of the school year. The positive or negative effect of the school site could be due to any of a number of mechanisms. The site might have a special program, a more motivated staff, more active parents, a better neighborhood, a better location, or other influences that make one school "better" than another. It is one of the strengths of the Capistrano analysis that we were able to include individual school sites as variables in the models to account for these effects. It is very noteworthy that the effect of moving from a classroom with the least to the most daylighting is of the same order of magnitude as the effect that would be seen by moving from an average school in the district to one of the highest, or lowest, performing schools in the district. Unverified absences had a slight negative impact on math improvement, but not on reading improvement. Ten unverified absences have the same order of magnitude effect (negative) as learning in a skylit or daylit room (positive). Size of school: The size of the school was found to have a small but significant negative effect. For the very largest school in the district, average student performance decreases by less than one percentage point. For the smallest, performance increases by about one half a percentage point. 2.1.4 Validity of the Model The Capistrano analysis was put through two additional statistical tests to determine the validity of the results. One test looked at the "explanatory power" of the daylight variables relative to the other variables included in the model. The daylight and window variables were relatively powerful when compared to the other variables, while the skylight and operable window variables tended to have lower explanatory power. However, in general, all the daylighting variables offered as good, if not a better explanation for how far a student would progress, as the variables for which school they attended, whether they were in a special language program, or how many absences they each had. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 17 August 20, 1999 The second statistical test ran the same data through a new model that looked at the average performance of each classroom group, rather than of individual students. The daylighting variables all remained highly significant in this test. This test implied that the influence of being in a given classroom group, whether because of the teacher or the class dynamics, was less significant than the variations between individual students. This may be because the Capistrano District does not group students into classrooms by abilities, or because the Capistrano teachers are all reasonably similar in their ability to teach the math and reading curriculum. However this statistical test did allay concerns that we had picked up a "teacher effect" instead of a "daylighting effect" in our analysis. 2.2 The Other Districts We performed a similar analysis for two other school districts, one in Seattle and another in Fort Collins. Due to limitations in the data, the analysis for these two districts was less detailed than for Capistrano. The studies in Seattle and Fort Collins used the absolute value of the students’ final scores on math and reading tests at the end of the school year, rather than the amount of change from the beginning of the year. As a result, more variables show up as significant in the models. For example, students’ ethnic background and socio-economic status become important predictors of their actual test scores, whereas in Capistrano these variables were not significant predictors of how far a student would progress in one year. We have less confidence in the results of these models, since the analysis was less detailed. There is more probability that there are other factors that we were not able to account for that could invalidate the results. However, we find it very suggestive that in two very different districts, in different states, we found very similar results to the Capistrano analysis. In both of these districts we also found large, positive, and highly significant effects for daylighting. 2.2.1 The Seattle District Seattle Public School District is a primarily urban school district in the city of Seattle, Washington. Its neighborhoods tend to be in the older, more densely settled areas of the city. It has also expanded by incorporating neighboring suburban districts. Elementary schools in Seattle tend to have far fewer students than Capistrano, and a great deal more floor space per student. Seattle provided us with student test score records for all elementary students at over 60 school locations. The test scores used in the analysis are from the Iowa Test of Basic Skills (ITBS), Form M, for grades 2 to 5, for math and reading, administered in the spring of 1998. In addition to the test scores, the data set included codes for the students’ classroom location, grade, ethnicity, gender, socio-economic status and participation in special programs. The elementary schools in Seattle have a large range of physical conditions. Mostly older, the schools range in age from 20 to 90 years old. Most are multiple story CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 18 August 20, 1999 buildings with interior hallways, and extensive indoor and covered facilities for student use, such as gymnasiums, covered play areas, libraries, cafeterias and auditoria. Many schools have had multiple additions over the years, but, in general, daylighting conditions within a given school were fairly similar across all classrooms. Most Seattle elementary schools have substantial windows with clear glass, although some have lightly tinted glass and a few have minimal or no windows. There are a few "open" schools from the 1970s with "pod" classrooms that share a common space in the center. These open classroom schools typically have few, if any, windows. Some schools are clearly designed for full daylighting, with high ceilings (11’) and window walls on two sides of the classroom. Daylight was also provided from clerestory windows high up in the walls, sawtoothed monitors or skylights in four of the schools. One school with open-type classrooms has high clerestory windows that allow daylight deep into the building. One group of classrooms has three small skylights along the inner wall, and another group has large central skylights with diffusing louvers that cover most of the ceiling. Figure 5: Older Seattle School With Windows Code 4, Exterior (left), Interior (right) Figure 6: Seattle Classrooms With Clerestory Windows (left) and Central Skylight and Diffusing Louvers (right) CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 19 August 20, 1999 2.2.2 The Fort Collins District The Poudre School District in Fort Collins, Colorado is a rapidly growing school district about two hours north of Denver, situated in the college town for Colorado State University. The district has many new facilities, some of which include aggressively daylit classrooms which are lit from rooftop windows, called sawtooth monitors. The sawtooth monitors in Fort Collins face south, and although they diffuse the sunlight somewhat, they are very bright. Teachers have the option of pulling an insulating shade across the skylight to darken the room. On one partly sunny winter day, we observed that 60% of classrooms had their shades closed. These skylit schools have modestly sized windows. Other older schools with the sawtooth monitors have somewhat larger window areas. However, none of the Fort Collins schools have classrooms with the very large windows designed for complete daylighting, as we found in the Capistrano or Seattle districts. The Fort Collins district provided us with data sets of student test scores for math and reading "level" tests for spring of 1998 for 23 schools. These level tests for math and reading, developed by Northwest Educational Association, are similar to the tests used in the Capistrano analysis. The data sets also included demographic information, similar to Seattle and Capistrano, including grade level, ethnicity, gender, socio-economic status, and special program codes. Based on an examination of district records, we added information to the database about the age and the size of the schools. We examined architectural plans for each school to determine classroom type (open vs. traditional classrooms), and develop the daylight, window and skylight codes. Due to the structure of the data sets given to us by Fort Collins, we were not able to identify students by their specific classroom location, but only by their grade level within a school. As a result the final analysis in Fort Collins was much simpler and more general than the other two districts. Luckily, most schools in Fort Collins had fairly uniform daylighting conditions for all their classrooms. Thus an overall school daylighting code was a reasonable approximation of individual classroom conditions. Figure 7: New Fort Collins School with Monitor Skylights CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 20 August 20, 1999 Figure 8: Interior of Fort Collins school with South Facing Monitor Skylight We were told that none of the schools in Fort Collins have air conditioning. Thus, information about air conditioning and natural ventilation was not included in our analysis for this district. All of the schools visited in Fort Collins had fluorescent lighting, but we could not confirm that fluorescent lighting was universal in all schools. Information about electric lighting was not included in our analysis for this district. All of the schools visited in Fort Collins had fluorescent lighting, but we could not confirm that fluorescent lighting was universal in all schools. Information about electric lighting was not included in our analysis for this district. 2.2.3 Seattle and Fort Collins Findings Both the Seattle and the Fort Collins analyses found a similar pattern of positive, significant results for the daylighting variables. These results were not only significant, but remarkably consistent in magnitude across all models. It should be remembered that these results are from different tests with different scales. The Seattle tests used a scale called normal curve equivalent which ranges from 1-99. The Fort Collins tests used the same RIT scale as Capistrano. We have put all the test results in our graphs on the same 1-99 scale in order to make the results between districts as comparable as possible. However, we are still trying to compare apples and oranges, so we must generalize and talk about fruit instead. The percentage effect is perhaps the best way to compare across districts. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 21 August 20, 1999 Seattle ITBS Iowa Test of Basic Skills NCE Scale 1-99 Spring Scores Reading Math Reading Math Reading Math Model 1 Daylight, Min. to Max. 7.5 5.6 99.9% 99.9% 13% 9% Model 2 Windows, Min. to Max. 7.7 8.7 99.9% 99.9% 13% 15% Skylights, Min. to Max. 3.9 3.4 99.9% 99.8% 7% 6% Difference in Average Test Scores (NCE percentage points) Statistical Certainty Difference as a % of District Average Score Analysis Results Percentage Effect Figure 9: Summary Daylight Findings for Seattle Figure 9 summarizes the percentage effects for the daylighting related variables of the four Seattle models. All these variables were found to have 99% certainty. All other things being equal, students in classrooms with the largest window area, or the most daylight, were found to be testing 9% to 15% higher than those students in classrooms with the least window area or daylighting. A 6% to 7% effect is observed for skylit classrooms. We do not report on a rate of improvement here because the Seattle models looked at the level of test scores at the end of the year, not the change between fall and spring, as in Capistrano. While the percentage effect is smaller, the magnitude of the difference in test scores is considerably larger in Seattle than Capistrano. This may be partially a function of a less detailed model. It may also reflect a cumulative effect of daylighting over a longer time period. For instance, if daylighting has a positive effect on learning, and if students stay at a well daylit school over the course of a few years, then the effects of daylighting might be cumulative over a student’s career, and thus larger than for a single school term. Fort Collins NEA Core Level Tests Normalized Scale 1-99 Spring Scores Reading Math Reading Math Reading Math Model 1 Daylight, Min. to Max. 3.8 3.4 99.9% 99.9% 7% 7% Model 2 Windows, Min. to Max. 10.2 7.0 99.9% 99.9% 18% 14% Skylight Monitor - 1.6 n/s 99.7% - 3% Difference in Average Test Scores (normalized RIT points) Statistical Certainty Difference as a % of District Average Score Analysis Results Percentage Effect Figure 10: Summary Daylight Findings for Fort Collins The Fort Collins results in Figure 10 show a 7% improvement in test scores in those classrooms with the most daylighting, and a 14% to 18% improvement for those students in the classrooms with the largest window areas. There is a 3% effect for math scores in the classrooms with the roof top monitors and no significant effect on reading scores. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 22 August 20, 1999 The Fort Collins results may be influenced by a number of factors which are distinctive about this district. First of all, we had the least amount of information about the characteristics of the students and schools in the Fort Collins district. Of the three districts studied there is the greatest likelihood that there may be other unknown variables that influenced the findings. Secondly, the district has only a modest range of window conditions. There were no classrooms in Fort Collins without any windows, and no classrooms with really large window areas, or what we considered "full" daylighting. Because of this limited range of window conditions in our model, the effect of going from minimum to maximum window area may be unreported. Finally, the skylighting variable is considerably weaker in these models than in Seattle, having only a small positive magnitude for math, and no significance for reading. We believe that the weak positive effect of the skylight variable may be a function of poor lighting quality from the south facing monitors, and the observation that many teachers seem to keep the shades down to solve this lighting quality problem. One would expect that skylights that are closed off much of the time would not have much of an effect. The results for the daylighting variable are probably also depressed for the same reason, since the daylighting code was a function of the skylighting code. We assigned the classrooms with skylights the highest daylight code for our analysis, on the expectation that they would have the highest daylight illumination levels. We didn’t know the extent of the glare problems or the operation of the shades until after the analysis was completed. Ideally, a daylight variable would be based on observations of daylight illumination conditions throughout the school year. Such observations, however, were beyond the resources of this study. 2.2.4 Other Variables The results for all the major variables of the Seattle regression models are presented below in Figure 11. There are many more variables than for Capistrano, as discussed above, since demographic variables remain important in predicting a student’s actual test level, rather than their yearly progress, as in Capistrano. We attempt to interpret these findings below. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 23 August 20, 1999 Seattle -12.00 -10.00 -8.00 -6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 Daylight Code 1-4 Window Code 1-4.5 Skylight Code 0-4.5 Portable Open Classroom Gifted room (70%+) School SF Students per Class School Pop +500 Ethnic 1 Ethnic 2 Ethnic 3 Ethnic 4 Gender Econ 1 Socio 1 Socio 2 Socio 3 B Coefficient ITBS Scores 1-99 NCE Reading Daylight Reading Skylight Math Daylight Math Skylight Figure 11: Seattle Relative Importance of Variables, Difference in Mean Score The gifted room variable has the greatest magnitude of effect. As would be expected, students in a classroom with many gifted children (70%+) are likely to score about 15 points higher than the mean. The school population variable shows a strong positive effect, so that the larger the school, the better students perform. This might seem to be contradictory to findings from other studies. However, given the small size of some Seattle schools, this may indicate that these schools are below an optimum size. Or it may be that larger schools in Seattle have some other advantages, such as better facilities. The demographic variables—ethnicity, economic and social status—are seen to have a strong influence. It is interesting to note, however, that the magnitude of these variables is mostly equal to, or less than, the daylighting variables. In later tests on the explanatory power of these variables it was found that the amount of daylight in a classroom was sometimes a more potent predictor for how well a student would perform than their gender, whether they were living in a single parent household, or how many students there were in their classroom. Other variables, such as the type of classrooms (portable or open), school area in square feet, and students per class, are seen to have occasional and modest impacts on student performance. CALIFORNIA BOARD FOR ENERGY EFFICIENCY CONDENSED REPORT DAYLIGHTING IN SCHOOLS HESCHONG MAHONE GROUP 24 August 20, 1999 Fort Collins -8.00 -6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 8.00 10.00 Daylight Code 1-5 Window Code 0-3 Skylight Code 0,1 Open Classroom Vintage 10 yrs School Pop +500 Ethnic 1 Ethnic 2 Ethnic 3 Ethnic 4 Gender Language Prog Socio 1 Socio 2 Socio 3 Econ 1 Econ 2 B Coefficient - Normalize RIT Points Difference from Mean Score Reading Daylight Reading Skylight Math Daylight Math Skylight Figure 12: Fort Collins Relative Importance of Variables, Difference in Mean Score The results of the Fort Collins regression equations in Figure 12 show a very similar pattern to Seattle. Indeed, the very similarity of the results for the diverse variables across districts argues for the validity of the models. It seems reasonable that there would be a change in the impacts of the ethnicity variables between the cities because of the different mix of immigrant populations in each. The daylight variables have about as large a positive effect upon the students as < |