Nurturing innovation: how much does collaborative
management help?
Dan Cooperman
Principal, Mountain View School, Santa Barbara, California, USA
The policy issue
Improving the academic performance of low-achieving students is a foremost concern of educators and policy makers. Toward meet-ing this goal, restructured school manage-ment and collaborative decision making have been proposed as measures which promote instructional practices that benefit under-achieving students (Fullan, 1991). Collabora-tive management can be defined as any school governance process for shared deci-sion-making among administrators, tea-chers, and parents on matters relating to instruction, personnel, student conduct, and budget. The underlying assumption is that instruction will improve because collabora-tive relationships empower and motivate site personnel to find instructional solutions for the educational problems that students face. Nonetheless, the linkage between collabora-tive management and innovacollabora-tive instruction for low-achieving students lacks extensive empirical evidence to validate the perceived connection (Sarason, 1990; Murphy, 1991; Elmore, 1992). It remains primarily specula-tion whether more democratic power rela-tionships actually engender instructional change at schools serving low-achieving students.
At the same time, there is growing concern for this group of students (Carnegie Forum on Education and the Economy, 1986), var-iously identified by the terms ``economically disadvantaged'', ``underserved'', ``low-achiev-ing'', or ``at-risk''. The consequences of under-education for the increasing numbers of these students affect larger society (Hodg-kinson, 1989) and include the emergence of a dual society with a large and poorly educated underclass, disruption to higher education, reduced economic competitiveness for in-dustries impacted by the growing number of workers with low educational attainment, and increased costs to provide public ser-vices to those living in poverty. The polar-ization of society which results from economic inequality intensifies political conflict and instability as well (Levin, 1990).
This article furthers current research by providing valuable empirical knowledge about the relationship between collaborative
management and innovative instruction in schools serving large populations of under-served students. By exploring the question of whether innovative instruction can be ``bred'' in a collaborative school environment, this research could inform policy makers and practitioners who seek to reform school by restructuring features of school organization and instruction. While not establishing a causal link between collaborative manage-ment and innovative practice, the research will reveal whether a statistically significant relationship exists between the two vari-ables. Organizational features and instruc-tional methods in elementary schools receiving supplemental federal funds through the Chapter I program are the focus of the research because these schools serve high populations of economically disadvan-taged, low-achieving students as a condition for participation in this compensatory pro-gram.
Background literature
Much of the research examines two back-ground factors which influence decisions affecting instructional practice. The first focuses upon individual teacher characteris-tics, such as educational attainment, years of teaching experience, or the degree of teacher involvement in professional development activities (Richardson, 1990; Little, 1981a). The second group of factors tend to be contextual, and include school size, the socio-economic level of the students, leadership of the principal, combined experience and sta-bility of the teaching staff, involvement with schoolwide in-service programs, or the tea-cher's sense of efficacy (Leeet al.,1993; Rowanet al., 1991; Newmannet al.,1989; Rosenholtzet al.,1986; Smylie, 1988; Leeet al., 1991; Richardson, 1990; Raudenbushet al., 1992; Friedkin and Slater, 1994).
As a means of influencing the outcome of instructional practice, two approaches to reform have been explored. The first ap-proach can be classified as systemic reform, that is, modifications of school leadership in order to improve schooling (McLaughlin and Talbert, 1990; McLaughlin, 1987). The central The International Journal of
Educational Management 13/3 [1999] 114±125
#MCB University Press [ISSN 0951-354X]
Keywords
Innovation, Management styles, Participative management, Schools, Teachers, Training
Abstract
question is whether a collaborative work environment can take teachers of varied background characteristics within diverse contextual settings and successfully nurture the conditions in which beneficial changes in instructional practice can take place. This approach is held to be most promising for fundamental and lasting improvement in educational practice (Elmore, 1993).
As an alternative to systemic reform, programmatic reform has emerged as a fundamentally different approach to school change, involving sustained inservice train-ing to implement innovation and improve instructional quality (Spady, 1988; Slavinet al.,1989). Examples of the instructional innovations proposed to assist low-achieving students include cooperative learning, con-tinuous progress learning programs, bilin-gual education (for students learning English as a second language), computer assisted instruction, and one-to-one tutoring. The purpose is to increase student motivation through stimulation of critical thinking skills and increased time in engaged learn-ing, that is, the time students spend learning academic content (Slavinet al.,1989; Hakuta, 1986). A repeated theme is the emphasis upon high expectations for student learning out-come when provided appropriate instruction (Stevenson and Stigler, 1992).
Other studies have suggested important relationships among the roles of organiza-tional design, teacher background, and school characteristics, and their effect upon teacher morale, sense of efficacy, and job satisfaction (Friedkin and Slater, 1994; Ro-wanet al.,1991). These studies also frame the need for empirical investigation into the connection between collaborative manage-ment and innovative instruction, and not just teacher perception and attitude. If teacher sense of efficacy can be affected by organiza-tional design, what precisely is the relation-ship of organizational design to teaching practice?
Moreover, current research leaves several questions unanswered. The direction of cau-sation between collaboration and self-efficacy is unclear; it could be the case that self-efficacy leads to greater inclinations for collaborative management. Second, it is not clear how strong the role of the principal is in bringing about change at the school site, as Rosenholtz (1985) suggests in her research on effective schools. Nor does research speak to the degree to which teacher training and experience blend with organizational fea-tures to produce greater inclination towards changing and improving teaching practice (Richardson, 1990).
The main limitation of empirical research in this area is the difficulty of linking instructional innovation to any one variable or sets of variables. There is a need to examine more carefully the relationship of management practice and classroom innova-tion, especially as it pertains to the instruc-tion of economically disadvantaged students. This interaction is especially timely in light of the policy and research effort being placed in the school decentralization and restruc-turing movements, two changes which em-phasize the importance of collaborative management features at the site level. The need for this research is particularly keen at the elementary school level. The urgency to make change is particularly great at low-income or Chapter I schools, where there is room for the greatest improvement in aca-demic performance.
Approach and method
These issues lead to the questions which will be studied in this research article. They are:
. What instructional practices do teachers
in Chapter I elementary schools report that they use?
. What are the organizational features of
Chapter I elementary schools as reported by their teachers and administrators?
. What effect do organizational features,
teacher background, and school charac-teristics have on the reported use of innovative instruction?
Hierarchical Linear Modeling (HLM) is used in this study to examine the linkages between school management, teacher characteristics, and instructional practice at the elementary school level. Since conventional statistical models, such as multiple regression, can only account for the variation of single-level observations, either at the individual or at the organizational level, the HLM statistical model is preferable because it allows us to examine the complex influences that school characteristics and teacher charac-teristics have upon instruction between in-dividual teachers within a school while also examining the mean instructional differ-ences among schools (Paterson, 1990; Bur-stein, 1980). Although some warn of
inferential problems which accompany use of HLM, there has yet to be found a more comprehensive alternative strategy (Rau-denbush and Bryk, 1986).
Data source
This study uses data from PROSPECTS, a research project conducted by Abt Associ-ates. Data were collected from 372 schools in Dan Cooperman
Nurturing innovation: how much does collaborative management help?
167 different districts serving 42,461 students. Cohorts of students in the first, third, and seventh grades were tracked over a six-year span beginning in 1991 and ending in 1996. Data collection consisted of a self-adminis-tered survey taken by district Chapter I coordinators, school principals, Chapter I teachers, classroom teachers and aides, bi-lingual and ESL teachers, parents, and stu-dents. Survey questions included school characteristics, teacher background, in-structional practice, school environment, and student academic performance. The data are distributed by census region ± northeast, south, midwest, and west ± and by urbaniza-tion category: urban, suburban, or rural. In addition, a qualitative study was conducted in 24 schools which employed innovative teaching strategies for underserved students (Stringfield, 1991).
The schools selected for this study were a subset of the 372 schools which reported data in 1991, and was restricted to those schools reporting English Teacher and Principal data for third grade. The third grade cohort was selected because teachers at this level generally employ a greater range of elementary school basic skill and critical thinking instruction than first grade tea-chers. The initial sample for the study consisted of 919 teachers, 207 principals, and 241 schools reporting third grade data for this year of the study. The final sample consisted of 669 teachers who had valid responses at 172 schools. Comparison be-tween the final sample and the initial sample, reported later in this section, did not reveal large differences.
Variables
Dependent variable
The construct of innovative instructional practice is measured by a nine-item scale covering such varied methods as cooperative learning, computer-assisted instruction, whole language approach to reading, teach-ing the writteach-ing process, one-to-one tutorteach-ing, language experience approaches, grouping practice, main approach to language arts instruction, and continuous progress learn-ing programs. These items closely corre-spond to the dimensions of effective instruction for economically disadvantaged learners outlined by Slavinet al.(1989), Meanset al.(1991), Hakuta (1986), and Bain and Herman (1990).
On the survey form from which these data were collected, teachers were to indicate whether they employed each of the nine innovative practices within their classrooms. The total number of affirmative responses
were tabulated for each teacher and recorded as a single composite score for innovative instructional practice. These measures will be aggregated for discussion at the individual and between-school levels. The means and overall distribution for this variable are found in Table I.
While over 50 percent of the teachers report that they use whole language teaching methods, the writing process, or cooperative learning strategies, there was less reported use of instructional methods such as group-ing students for reasons other than ability level, the use of computers, and the language experience approach to teaching reading. These latter methods allow students of varied skill levels to work together to ensure equitable access to the core curriculum while improving the academic performance of at-risk students.
The outcome variable, Innovative Instruc-tion, concludes that teachers in the data set employ on average a combination of roughly three of the nine possible methodologies which current research suggests to promote the performance of low-achieving students.
Independent variables
The outcome measure of instructional prac-tice depends heavily upon individual, orga-nizational, and contextual factors. Individual background can affect the method used by that teacher, just as the socio-economic level of students or school management features may contribute to the choice of teaching methodology. In order to test these ideas, several independent variables are used, some at the individual level and some at the school level.
A first set of variables refers to the back-ground characteristics of teachers that could affect their instructional approach. Previous research has shown that these variables include years of experience (number of years taught), educational level (bachelor's degree, Master's, or advanced training beyond the Master's degree), race, belief in students' ability to learn, and involvement with pro-fessional development activities (Richard-son, 1990).
The second set deals with measurement of school characteristics. These variables in-clude school size (total number of students enrolled), the SES level (as measured by percentage of students receiving the federal free lunch program), the percentage minority (the percentage of Black and Latino students enrolled) and sector (urban, suburban, or rural).
Additional school-level variables were identified from studies which address school organizational features (Newmannet al., 1989; Rowanet al.,1991; Raudenbushet al., 1992; and Leeet al., 1991). Research suggests Dan Cooperman
Nurturing innovation: how much does collaborative management help?
considerable variation between schools in management design (Rosenholtz, 1985), and these variables measure the degree to which collaborative management features are im-plemented within a school. These features are: staff influence over school policy, op-portunities to participate in decision making (on curriculum, discipline policy, and pro-gram), level of staff collegiality (time spent working together, uniformity of goals and beliefs, and willingness of teachers to help one another), and parental influence on school policy.
Two additional aggregate variables were constructed to examine environmental factors at the school level. These variables recorded whether an individual school had more than one-half of its teachers report a strong belief that remedial students could learn given appropriate instruction, and whether more than one-half the teachers at the site had been involved in an in-service program for more than 35 hours of instruc-tion during the past year. Those schools which had more than 50 per cent agreement or participation levels were recorded as a value of one within two dummy variables, each representing the presence of these features at a school.
The results of these scales will be reported at two levels of analysis: the individual and between-school levels. Figure 1 diagrams the variables and their interactions. Descriptive statistics for the individual variables are presented in Tables II-IV.
Descriptive differences among variables On the school-level, an interesting contrast is found in the reports for participatory deci-sion making between the principal survey
and the teachers' survey. Seventy-eight per-cent of principals expressed strong agree-ment that staff and administration worked well together while only 24 percent of tea-chers reported strong agreement that princi-pals consulted them concerning decisions affecting teachers. Furthermore, only 32 percent of teachers strongly agreed that high levels of collegiality are found at their school sites. The contrast shows a stark difference in perception based upon job title.
Moreover, it is noteworthy that less than 10 percent of parents were reported to be involved in making policy decisions. Despite principal reports of good working relations at the school sites, teachers and parents are either less involved or less satisfied with their involvement in decision-making pro-cess or collegial relationships.
Finally, few schools have widespread staff involvement with sustained inservice pro-grams or belief in the ability of remedial students to learn. Only 15 percent of schools have staffs in which over half the teachers report strong belief in student learning ability, and only 19 percent of schools have over one-half the staff involved with inser-vice training of more than 35 hours in the past year.
Sample bias
Sample bias is reduced in the study because schools and districts were required to parti-cipate as a condition of Chapter I funding. Nonetheless, missing data were higher for teachers and schools in areas of the greatest poverty or lowest achievement. To provide a descriptive view of possible response bias, Table V presents the results of a comparison of means for the study sample and for schools
Table I
Mean and standard deviations for the components of the innovative instruction composite variable
Variable N Mean SD
Non-ability based groupings 919 0.06 0.23
Language experience approach for reading 919 0.16 0.37
Computers to teach reading 919 0.16 0.37
Programmed instructional materials to teach reading 919 0.32 0.47
Whole language or language experience methods as main instructional method for reading
919 0.40 0.49
Certificated tutors for individualized instruction 919 0.42 0.49
Whole language approach to teach reading 919 0.54 0.50
Writing process to teach language arts 919 0.59 0.49
Cooperative groupings for reading activities 919 0.62 0.49
Outcome variable
Composite, 0-9 scale, for use of innovative practices listed above 919 3.01 1.84
Note:Instructional variables are dummy variables indicating teachers who reported that they use or frequently use, this practice. Consequently, the means represent the proportions of teachers who responded to each item. Outcome variable is a composite variable
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
excluded from the analysis due to incomplete information in the data files at the individual and school levels.
The means for the sample group and schools dropped from the sample are quite similar, with two exceptions. There is a modest variation in the mean for educational attainment for teachers in participating schools and those not included in this study. This difference, statistically significant at the 0.10 level, may reflect the interest of teachers with Master's or doctoral degrees for taking part in studies which could improve school-ing. Another discrepancy is the mean for the standardized composite for instructional in-novation. Although not statistically signifi-cant, the disparity between groups for this variable may show that schools which parti-cipate in research are in general more disposed toward changes in instructional practice, or reflective analysis which could generate instructional change. Whatever the
cause, it can be speculated that greater variation in the outcome variable could have been registered had more schools fully com-pleted the PROSPECTS survey.
Statistical models
Statistical models used in this study are discussed in the Appendix.
Results
Regression results
Linear regression is used to assess the significance of teacher background charac-teristics on instructional practice, and to determine the degree to which the teacher level model relates to the use of innovative instructional measures. Through multivari-ate analysis, only three variables produce T-scores of great significance. These vari-ables were: nine years' or less teaching experience, 35 hours or more of in-service training in the past year, and holding a strong belief that remedial students can learn given appropriate instruction. The positive effect for each of these three variables was strengthened when analysis was taken from the univariate to the multivariate level, as demonstrated by the increased coefficient for each of the variables shown for the two levels of estimates (see Table VI).
While three variables show significance in estimating the predicted instructional prac-tice at the teacher-level, it is noteworthy that the teacher-level model accounts for very little of the variation in teaching method among teachers. The adjustedR2value for the multivariate estimate was 0.054, meaning that a large measure of variation ± over 90 percent ± is unexplained by this model. The purpose of this study, however, is to examine variables which administrators and policy makers can manipulate to influence the use of innovative teaching methodologies at the school level. For this reason, the HLM models which follow are of greater importance, for they show the degree to which instruction can be affected by variables at the school level.
Analytic models
The outcome measure considered in the HLM analyses is instructional practice. The tea-cher level model regresses instructional practice as a function of years of teaching experience, teacher ethnicity, teacher educa-tional attainment, time spent in recent in-service training, and the teacher's belief in the ability of remedial students to learn given appropriate instruction. Between school dif-ferences are characterized in terms of six
Figure 1
Conceptual framework for analyzing teacher characteristics, school characteristics, organizational features and instructional practice
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
models to estimate the distribution of in-structional practice according to the influ-ences of individual teacher characteristics, school features, student characteristics, ag-gregated teacher characteristics, and school organization. The output for Models I, II, V, and VI are shown because of their signifi-cance to the results of the study.
HLM results
Model I: Examining variability among schools
The first HLM model tests the hypothesis that innovative instructional practice varies across schools. This model is the null model. It permits estimation of the mean level of innovation across schools.
Using the standardized outcome variable for instructional practice, a mean coefficient of 0.039 is observed for all schools. More important is the conclusion that differences in instructional practice can be found among schools, and that approximately 22 percent of this variation can be accounted for by differences between schools in contrast to 78 percent of the variation accounted for by differences among individual teachers. Table VII shows the results for Model I.
Model II: The teacher level (Level I) model Having found in Model I that mean instruc-tional practice differs among schools, the remaining models attempt to explain these differences. Model II asks the extent to which these differences can be accounted for by individual teacher background characteris-tics. Within this model, the estimated coeffi-cients for several variables exceed the grand mean for instructional practice. Belief in the ability of remedial students to learn, in-service programs of six hours or more per year, attainment of a Master's degree or higher, and less than ten years' teaching experience all tend to yield higher average levels of innovative practice. At the same time, only two variables are statistically significant in their effect:
1 Belief in the ability of remedial students to learn with appropriate instruction (p= 0.032); and
2 In-service training of 35 hours or more in the past year (p= 0.004).
While these variables account for some of the variance in instruction at the teacher level, more variability remains to be explained. The results of the Teacher Level Model are given in Table VIII.
Model V: The aggregated teacher characteristics model
The next step involved modeling of the distribution of instructional practice as a
Table II
Means, standard deviations and descriptions of teacher-level variables
Variable N Mean SD Type Description
Demographic
Minority 655 0.22 0.41 D 1 = Minority
background Experience
1-9 yrs 632 0.34 0.47 D
10-20 yrs 632 0.41 0.49 D
21-38 yrs 632 0.25 0.44 D
Education
Academic attainment 659 0.34 0.47 D 1 = Master's
degree or higher Belief in remedial
students' ability to learn 635 0.15 0.36 D 1 = Strong belief
In-service
0-6 hrs 626 0.11 0.31 D
6-35 hrs 626 0.68 0.47 D
35+ hrs 626 0.21 0.41 D
Teacher outcome variables
Innovative instruction 669 0.02 1.00 C Standardized
composite of instruction
Note: Means represent proportions of teachers in each category. Variable type is: dummy (D) or composite (C)
Table III
Means, standard deviations and descriptions of school-level variables
Variable N Mean SD Type
School enrollment
325 or less 172 0.28 0.45 D
326-466 172 0.31 0.47 D
467-675 172 0.21 0.40 D
676-1730 172 0.20 0.40 D
Location
Urban 172 0.42 0.50 D
Suburb 172 0.13 0.34 D
Rural 172 0.45 0.50 D
Percent students in free lunch program
0-25 172 0.27 0.45 D
26-50 172 0.22 0.41 D
51-75 172 0.24 0.43 D
76-100 172 0.27 0.44 D
Percent white student composition
0-25 172 0.26 0.44 D
26-50 172 0.14 0.35 D
51-75 172 0.11 0.31 D
76-100 172 0.49 0.50 D
Note:Means represent proportions for each category. Variable type is dummy (D) or composite (C)
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
function of two teacher characteristics ag-gregated at the school site:
1 belief in the ability of remedial students to learn with appropriate instruction; and 2 participation in in-service training for 35
hours or more during the past year.
These variables were selected because of their significance at the individual teacher level. In turn, schools at which more than 50 percent of the third grade staff responded affirmatively to these dummy variables were selected to estimate the effect of these para-meters on instructional practice.
The relative effects of these variables were quite strong. The coefficients for instruction (Belief = 0.411; In-service = 0.398) were the highest for any of the variables in the study. In addition, each variable was statistically significant at the 0.01 level. As will be seen later in the results for variance by each model, Model V also explains almost one-fifth of the variance in instructional practice between schools. This model accounted for more of the variability between schools than any other model. Results for Model V are given in Table IX.
Model VI: The school organization model As discussed earlier, collaborative manage-ment or participatory managemanage-ment struc-tures have been postulated as reforms which could breed instructional innovation. Model VI estimates the effects of variables pertain-ing to these elements on instructional meth-ods. The results support the views of
researchers Sarason (1990), Weiss (1993), and Newmann (1993), in that statistical
significance cannot be found among the key variables concerning collaborative manage-ment. Among the variables, only parent participation in policy decisions and
Table IV
Means, standard deviations, and descriptions of school-level variables
Variable N Mean SD Type Description
Organizational features Parents participate in policy
decisions
172 0.09 0.29 D 1 = strongly
agree Staff and principal work
well together
172 0.78 0.42 D 1 = strongly
agree Principal consults staff for
important decisions
172 0.24 0.31 D 1 = strongly
agree
High staff cooperation 172 0.32 0.30 D 1 = strongly
agree High schoolwide belief in
remedial
student ability to learn
172 0.15 0.35 D 1 = 50 percent or higher, staff belief in remedial students' ability to learn 50 percent or higher, staff
participation in in-service program, 35+ hours, past year
172 0.19 0.39 D
Note:Means represent proportions for each category. Variable type is dummy (D) or composite (C)
Table V
Means and standard deviations for schools in final sample and for schools excluded
Final sample Excluded sample
(N = 173) (N = 68)
Variable Mean SD Mean SD
Experience 1-9 yrs 0.305 0.265 0.341 0.326
Experience 10-20 yrs 0.391 0.279 0.371 0.332
Experience 21-38 yrs 0.258 0.261 0.233 0.315
Education 0.452 0.538 0.324* 0.308
Minority background 0.382 0.801 0.357 0.686
Innovative instruction 0.055 0.712 0.003 0.717
Note:*Significant difference between groups at the 0.10 level
Table VI
Regression analysis: teacher characteristics and instructional practice
Univariate estimates Multivariate estimates
Variable Coefficient T Sig. T Coefficient T Sig.T
Experience 1-9 yrs 0.197 2.77 0.006 0.251 3.12 0.002
Experience 21-38 yrs ±0.018 ±0.230 0.818 0.073 0.836 0.404
Education 0.070 1.00 0.316 0.102 1.37 0.171
In-service, 6-35 hrs ±0.227 ±3.14 0.002 0.195 1.73 0.084
In-service, 35+ hrs 0.463 5.73 0.000 0.632 5.00 0.000
Belief in remedial student ability 0.240 2.64 0.009 0.255 2.76 0.006
Minority background 0.064 0.801 0.423 0.040 0.480 0.632
Note:**Multivariate adjusted R square: 0.054
Table VII
HLM results for Model I: Null model
Fixed effect Coefficient Standard error
Average innovation mean, é00
0.039 0.051
Random effect Variance component
df pvalue
Level 1 effect,rij 0.788
School mean,uoj 0.225 171 0.000
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
principal consultation of staff on decisions affecting them approach statistical signifi-cance (parents,p= 0.130, consult,p= 0.142), but neither achieves this designation at the 0.10 level. Moreover, variation among schools cannot be explained using HLM estimates for this model. Results for model VI are given in Table X.
Explained parameter variance
Table XI presents the central findings of the study. It shows the estimated parameter variances for the individual and school distributive effects for Models II through VI.
The Table shows the substantial propor-tion of variance explained at the school level for the aggregated teacher characteristics model (19 percent). The teacher characteris-tics model, school features, and school orga-nization models each account for 3 percent of the variability at the teacher level, while
student characteristics and aggregated tea-cher characteristics account for 2 percent of the variation at the teacher level.
Among the models, only school features and aggregated teacher characteristics have any effect on between school variability. The comparatively large influence of aggregated teacher characteristics noted above is much greater than the 2 percent effect noted for school features. In conclusion, the aggregated teacher characteristics model is the best for explaining difference for schoolwide instruc-tional practice. Causation is not proven, however. Estimated school effects may be a function of teachers working at the school site, and actually reflect unidentified differ-ences among teachers working at those schools (Friedkin and Slater, 1994). Other potential limitations and the implications of these results for practitioners and research-ers are discussed below.
Discussion and conclusions
Summary of findings
Two surprising elements emerged from this study which hold importance to the study of collaborative management and innovative teaching practice. The first is that individual teacher background characteristics ± such as years of teaching experience, educational attainment, or ethnicity ± fail to produce significant results in explaining the intro-duction of innovative practice among indivi-dual teachers or differences in instruction among schools. The linear regression study demonstrates that the majority of the differ-ence in the use of innovative practice among teachers is associated with characteristics other than the background variables used in this study. These results suggest that while approximately three-quarters of the differ-ence among teachers in the use of innovative practice can be explained at the teacher-level, another explanatory model should be devel-oped and should reach beyond intuitively appealing variables such as experience, training, and teacher ethnicity.
A second finding is that school character-istics and organizational features examined in this study may be less important to explaining variations in the use of innovative practice than school reformers may believe. In this study, little variation was explained between schools by factors such as school size, the socio-economic level of student families, collegial staff relations, a
principal's consultation with staff concern-ing decisions affectconcern-ing them, or parental participation in policy decisions. This
Table VIII
HLM results for Model II: Teacher-level
Fixed effect Coefficient Standard error p value
Intercept 2, é00 0.041 0.051 0.427
Experience, 1-9 yrs, û1 0.134 0.086 0.120
Experience, 21-38 yrs, û2 ±0.001 0.093 0.991
Education, û3 0.087 0.081 0.284
Inservice, 6-35 hrs, û4 0.228 0.157 0.146
Inservice, 35+ hrs, û5 0.525 0.180 0.004*
Belief in remedial students' ability to learn û6
0.244 0.114 0.032*
Minority, û7 0.037 0.100 0.709
Final estimation of variance components:
Variance component df
Level 1,r 0.768
Level 2,uoj 0.230 171
Note:Level 2 variance is almost the same as the null model
Table IX
HLM results for Model V: Aggregrate teacher characteristics
Fixed effect Coefficient SE p value
Intercept 2, é00 ±0.086 0.057 0.131
Schoolwide in-service, 35+ hours é01
0.398 0.128 0.002*
Schoolwide belief in remedial students' ability to learn, é02
0.441 0.149 0.003*
Final estimation of variance components:
Variance component df
Level 1,r 0.774
Level 2,uoj 0.183 169
Note:Fixed effects for In-service, 35+ hours and belief in remedial students' ability to learn not shown in above Table
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
finding tends to confirm the work of Sarason (1990) and Elmore (1993).
Without a doubt, the major conclusion of this study is that two schoolwide environ-mental factors account for nearly one-fifth of the variation in the use of innovative prac-tice between schools. The first element is having a majority of staff members who express a strong belief in the ability of remedial students to learn given appropriate instruction. This outcome makes sense. Po-sitive attitudes among teachers most likely lead to positive actions which help children learn. This conclusion is supported by re-searchers such as Comer (1990) who assert that the quality of relationships and attitudes among school personnel and students are powerful factors in determining educational outcomes for low achieving students.
The second element bearing statistical significance was having greater than 50 percent of staff members participate in an inservice program which provides more than 35 hours of training per year. As this level of involvement and time commitment is achieved, statistically significant changes in instruction are observed. To this extent, the conclusions of researchers Richardson (1990) and Little (1981b) are supported by this study.
These researchers identify a key element ± inservice training ± that may yield positive effects upon a teacher's use of innovative instructional practices.
Implications for research
Several implications for further research can be identified. This study constructed an instructional variable based upon innovative strategies for teaching language arts (listen-ing, speak(listen-ing, read(listen-ing, and writing), but it would be of interest to examine the outcome of a study in which innovative instruction characteristics were related to teaching mathematics. Another subject area may be affected differently by collaboration or tea-cher background characteristics. A second area would be to examine in greater depth the effects of parent involvement in policy decisions, or for principal consultation with staff. These two variables came close to achieving statistical significance in the HLM analysis. Perhaps a more detailed examina-tion of these features could uncover a rela-tionship to innovation untouched by this study.
Another topic for exploration would be a similar study using a different sample. Per-haps schools which serve students who come from more advantaged backgrounds operate differently, and these differences may affect how teachers teach. Furthermore, the data set used for this study was not necessarily intended for the analysis conducted, and it relies upon self-reported data. It is possible that a survey designed uniquely for this study purpose would have produced more significant relationships among teacher background, school features, and innovative practice. While cross-checking was done, in some cases the outcome for a specific vari-able may represent a self-affirming assess-ment of one's own professional practice rather than an accurate reflection of man-agement style or classroom practice (McDonnell and Hill, 1993). A qualitative study could help to validate or disconfirm the self-assessments in this survey data. Through sustained observation, interviews, and other data, a qualitative researcher may find results which lend greater support to the value of collaborative management.
Finally, an area unexamined by this study is the linkage of school features, teacher characteristics, and innovative teaching practice to student performance. Ultimately, the case for collaborative management will need to be supported by improved student learning outcomes. This study focused upon the connection between school and teacher variables and innovative practice because there was a gap in research evidence making
Table X
HLM results for Model VI: School organization
Fixed effect Coefficient SE p value
Intercept 2, é00 ±0.129 0.123 0.292
Parent participation in policy decisions, é01 0.262 0.173 0.130
Principal and teachers work well together, é02 0.038 0.124 0.763
Staff collegiality, é03 0.157 0.194 0.419
Principal consults staff about important decisions affecting staff, é04
0.283 0.192 0.142
Final estimation of variance components:
Variance component df
Level 1,r 0.765
Level 2,uoj 0.231 167
Note:Fixed effects for in-service, 35+ hrs and belief in remedial students' ability to learn not shown in above Table
Table XI
Summary of results for variance explained by hierarchical linear models
Level 1 variance Level 2 variance Parameter
estimate
Variance explained
Parameter estimate
Variance explained
Innovative instructional practice
Teacher-level model (Level 1) 0.768 3% 0.230 ±
School features model 0.768 3% 0.220 2%
Student characteristics model 0.769 2% 0.240 ±
Aggregated teacher characteristics model 0.774 2% 0.183 19%*
School organization model 0.765 3% 0.231 ±
Note:*Significant at 0.005 level
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
this first important connection. Clearly, one more step is needed to better understand the interactive effects of collaborative manage-ment, effective innovative practices, and student learning.
While these suggestions would further exploration of the issues identified by this study, a cautionary note is to be made that the linkages identified through HLM analysis are associative rather than causative. Although the degree to which the variables are associated may be found, this measure-ment alone fails to establish a causal link among them, for other unseen factors may also be responsible for the results obtained.
A second consideration is that cross-sec-tional data, such as the data set in this study, examine the complexity of collaborative management at a single point in time. An argument can be made that the effects of collaborative management should be studied over a greater span of time. While a different method could offer certain advantages, the data used for this research permits the use of HLM, a method which uniquely provides analysis of the multi-level effects that indi-vidual teachers and schools have upon in-struction.
Conclusion
It is generally unwise to base educational policy or practice upon the outcome of one study. Nonetheless, the effort to analyze research results can lead to important new insights which in turn can help piece together the puzzle of effective school reform. In this respect, this study poses a good deal of hope for the future. The outcomes identify two specific features of schooling which account for nearly one-fifth of the difference in the use of innovative teaching practice between different schools. Given the enor-mous need to provide better schooling for low-achieving students, we could easily em-phasize the factors associated with positive results. For example, only 15 percent of schools in this study registered strong staff belief in the ability of all students to learn, or only 20 percent are involved in in-service programs of sufficient duration to truly effect instructional change. Better results could be obtained with greater policy support of funding for sustained inservice programs and refinements in teacher recruitment and selection to better screen for belief in student learning ability.
The most promising of the other policy implications rests in the findings of results for variables relating to site-based decision making for teachers and parents. While less significant than in-service and collective staff belief in student ability to learn, the study
supports further exploration of these prac-tices at the policy level. Such policies could be designed at the local level, and could be reflected in the guidelines that school boards give to administrators with respect to job responsibilities, or the importance given to such school-based decision-making bodies as a school site council. There is cause for optimism in these observations. If intensified focus on specific measures leads to greater use of innovative instruction for low achiev-ing students, then this optimism is justified. This study, and others which may follow, can lead to forward progress in the struggle to raise educational outcomes for underserved students.
References
Bain, J.G. and Herman, J.L. (1990),Making Schools Work for Underachieving Minority
Students, Greenwood Press, New York, NY.
Bryk, A.S. and Raudenbush, S.W. (1992), Hier-archical Linear Models: Applications and Data
Analysis Methods, Sage Publications,
New-bury Park, CA.
Burstein, L. (1980), ``The analysis of multi-level data in educational research and evaluation'',
Review of Research in Education, Vol. 8,
pp. 158-233.
Carnegie Forum on Education and the Economy (1986),A Nation Prepared: Teachers for the
21st Century,The report of the task force on
teaching as a profession, New York, NY. Comer, J. (1990), ``Building quality relationships'',
in Bain, J. and Herman, J. (Eds),Making Schools Work for Underachieving Minority
Students,Greenwood Press, New York, NY.
Elmore, R.F. (1992), ``Why restructuring alone won't change teaching'',Educational Leader-ship, April, pp. 44-8.
Elmore, R.F. (1993), ``School decentralization: who gains? Who loses?'', in Hannaway, J. and Carnoy, M. (Eds),Decentralization and School
Improvement, Jossey-Bass, San Francisco,
CA.
Friedkin, N.E. and Slater, M.R. (1994), ``School leadership and performance: a social network approach'',Sociology of Education,Vol. 67 No. 2, April.
Fullan, M.G. (1991),The New Meaning of
Educa-tional Change,Teachers College Press, New
York, NY.
Hakuta, K. (1986),Mirror of Language: The Debate
on Bilingualism,Basic Books, New York, NY.
Hodgkinson, H.L. (1989),The Same Client: The Demographics of Education and Service
De-livery Systems,Institute for Educational
Lea-dership, Washington, D.C.
Lee, V.E., Dedrick, R.F. and Smith, J.B. (1991), ``The effect of the social organization of schools on teachers' efficacy and satisfac-tion'',Sociology of Education,Vol. 64, July, pp. 190-208.
Lee, V.E., Smith, J.B. and Cioci, M. (1993), ``Teachers and principals: gender-related
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
perceptions of leadership and power in secondary schools'',Educational Evaluation
and Policy Analysis,Vol. 15 No. 2, pp. 153-80.
Levin, H.M. (1990), ``The educationally disadvan-taged are still among us'', in Bain, J. and Herman, J. (Eds),Making Schools Work for
Underachieving Minority Pupils, Greenwood
Press, New York, NY, pp. 3-12.
Little, J.W. (1981a),School Success and Staff Development: The Role of Staff Development in Urban Desegregated Schools, Executive
Sum-mary,Center for Action Research, Boulder,
CO.
Little, J.W. (1981b),The Power of Organizational Setting: School Norms and Staff Development, National Institute of Education, Washington DC.
McDonnell, L. and Hill, P.T. (1993), ``Immigrant education: the incredible shrinking priority'',
Education Digest,Vol. 59 No. 3.
McLaughlin, M.W. (1987), ``Learning from experi-ence: lessons from policy implementation'', Educational Evaluation and Policy Analysis, Vol. 9 No. 2, Summer, pp. 171-8.
McLaughlin, M.W. and Talbert, J.E. (Eds) (1990), The Context of Teaching in Secondary Schools:
Teachers' Realities, Teachers College Press,
Columbia University, New York, NY. Means, B., Chelemer, C. and Knapp, M.S. (1991),
Teaching Advanced Skills to At-Risk Students, Jossey-Bass Publishers, San Francisco, CA. Murphy, J. (1991),Restructuring Schools:
Captur-ing and AssessCaptur-ing the Phenomena,Teachers
College Press, New York, NY.
Newmann, F.W. (1993), ``Beyond common sense in educational restructuring: the issues of con-tent and linkage'',Educational Researcher, Vol. 22 No. 2, pp. 4-13, 22.
Newmann, F.W., Rutter, R.A. and Smith, M.S. (1989), ``Organizational factors that affect school sense of efficacy, community and expectations'',Sociology of Education,Vol. 62, October, pp. 221-38.
Paterson, L. (1990), in Raudenbush, S. and Willig, J.D. (Eds),Schools, Classrooms, and Pupils: International Studies of Schooling from a
Multicultural Perspective, Academic Press,
New York, NY.
Raudenbush, S. and Bryk, A. (1986), ``A hierarch-ical model for studying school effects'',
So-ciology of Education,Vol. 59, pp. 1-17.
Raudenbush, S.W., Rowan, B. and Cheong, Y.F. (1992), ``Contextual effects on the self-per-ceived efficacy of high school teachers'',
Sociology of Education,Vol. 65, April,
pp. 150-67.
Richardson, V. (1990), ``Significant and worth-while change in teaching practice'',
Educa-tional Researcher,Vol. 19 No. 7, October,
pp. 10-18.
Rosenholtz, S.J. (1985), ``Effective schools: inter-preting the evidence'',American Journal of
Education,pp. 353-88.
Rosenholtz, S.J., Bassler, O. and Hoover-Dempsey, K. (1986), ``Organizational
conditions of teacher learning'',Teaching and
Teacher Education,Vol. 2 No. 2, pp. 91-104.
Rowan, B., Raudenbush, S.W. and Kang, S.J. (1991), ``Organizational design in high schools: a multilevel analysis'',American
Journal of Education,pp. 238-66.
Sarason, S.B. (1990),The Predictable Failure of School Reform: Can We Change Course before
It's Too Late?,Jossey-Bass Publishers.
Slavin, R.E., Karweit, N.L. and Madden, N.A. (1989).Effective Programs or Students at Risk, Allyn & Bacon, Needham Heights, MA. Smylie, M.A. (1988), ``The enhancement function
of staff development: organizational and psy-chological antecedents to individual teacher change'',American Educational Research
Journal,Vol. 25 No. 1, pp. 1-30.
Spady, W.G. (1988), ``Organizing for results: the basis of authentic restructuring and reform'',
Educational Leadership,October, pp. 4-8.
Stevenson, H.W. and Stigler, J.W. (1992),The Learning Gap: Why Our Schools Are Failing and What We Can Learn from Japanese and
Chinese Education, Summit Books, New York,
NY.
Stringfield, S. (1991), ``Implementing a research-based model of Chapter I improvement,Phi
Delta Kappan, Vol. 72 No. 8, April.
Weiss, C.H. (1993), ``Shared decision making about what? A comparison of schools without teacher participation'',Teachers College
Record,Vol. 95 No. 1, Fall.
Appendix
Statistical models
This study employed hierarchical linear models to explain variation in instructional practice at two levels of analysis. The first equation estimates the relationship between individual teacher background and instruc-tional practice within each school, while the second set of equations estimate the rela-tionship between school characteristics and the mean level of instructional practice across schools. The purpose is to examine the degree to which teacher background, school organizational design, and school character-istics can be predictive of instructional practice at the individual (teacher) and group (school) levels. HLM estimates are generally derived through a series of equations. The first set of equations are referred to as teacher-level models.
{INNOVATIVE INSTRUCTIONAL PRAC-TICE}ij=ûoj+rij
Where INSTRUCTIONAL PRACTICEijis the instructional score for teacheriin school j, ûojis the mean score for schoolj,andrijis a random error term for teacheriin schoolj.
This equation allows the HLM program to partition the total variance in instructional practice at the individual and between school Dan Cooperman
Nurturing innovation: how much does collaborative management help?
levels. The second within-school individual model is:
{INSTRUCTIONAL PRACTICE}ij=Boj+ Bj(teacher characteristics)ij+rij
Where INSTRUCTIONAL PRACTICEijis the instructional score for teacheriin school j, Bojis the mean score for schoolj, Bjis a vector of coefficients measuring the effect of an array of teacher characteristics on in-structional practice at schoolj, andrijis a random error term for teacheriin schoolj.
The individual teacher characteristic variables are:
. û1= one to nine years' teaching
experi-ence
. û2= 21 to 38 years' teaching experience . û3= Master's Degree or Doctorate . û4= Teacher of minority background . û5= High belief in the ability of remedial
students to learn given appropriate instruction
. û6= In-service, 6-35 hours within the past
year
. û7= Inservice, 35 hours or more within
the past year.
These variables were grand mean centered with the exception of the variables for in-service and belief in the ability of remedial students to learn given appropriate instruc-tion. Grand mean centering adjusts the mean level of innovative instructional practice for each school by assuming that all schools had teachers with similar background character-istics ± experience, advanced degrees, and minority status, for example. Adjusted means provide a better way to estimate the effects of school-level variables on mean levels of innovative instruction by controlling for teacher characteristics that may also affect innovative instruction, but which schools cannot control.
The variables for in-service training and belief were group mean centered so that mean levels of innovation across schools were not changed because of differences in the attributes of teachers across schools. The school-level model is:
ûoj= é00+Uoj
where the sample mean for teacher instruc-tional practice is equal to the grand mean of all schools plus an error term for all schools.
The second between-school model esti-mated the effects of school characteristics on mean levels of innovative instruction among schools. Although HLM can be used to estimate the effects of school-level variables
on the other within-school variables (û1j, û2j, etc.), in this study there were too few teachers in each school to reliably estimate other school-level equations. Thus, this study focused on simply estimating mean outcomes for each school, known as means-as-outcomes models (Bryk and Raudenbush, 1992).
ûoj= é0(sample mean of instructional practice) + é1(measure of management features at the school site)j+ é2(other school level characteristics)j+Uj(Error)
Where éois the sample school mean for instructional practice, é1is a vector of coefficients which estimate the degree to which collaborative management practices are used at the school site, é2is a measure of other school characteristics which may in-fluence instruction, andUjis a random error term for schoolj.
In the between-school model, forûoj man-agement features and school level character-istics are entered in various analytic categories which specify the exact variables indicated in the between-school model given above. The equations for each model are below. The models were estimated in a series of steps. The first model was tested, after which only significant coefficients were retained for the test of the next model.
School features model
ûoj= é00+ é01(school size, 325 students or less) + é02(school size, 676 to 1,730 students) + é03(urban school) + é04(rural school) +Uj
Student characteristics model
ûoj= é00+ é01(0-25 percent of students on free lunch) + é02(76-100 percent of students on free lunch) + é03(white student population 0-25 percent of total) + é04(white student population 76-100 percent of total) +Uj
Aggregated teacher characteristics model ûoj= é00+ é01(schoolwide, 50 percent or more of teachers participate in 35 or more hours of in-service, past year) + é02 (school-wide, 50 percent or more of teachers believe strongly that remedial students can learn given appropriate instruction) +Uj
School organization model
Boj= é00+ é01(parents participate in making policy decisions) + é02(principal and tea-chers work well together) + é03(high level of staff collegiality) + é04(principal consults staff concerning decisions affecting teachers) +Uj
Dan Cooperman
Nurturing innovation: how much does collaborative management help?
