Reading First in Massachusetts Review of Student Assessment Data Presented Online April 13, 2009 Jennifer R. Gordon, M.P.P. Research Manager
Questions Addressed Today Have student assessment results in participating schools improved over time? Is there evidence that RF is closing the performance gap for key demographic subgroups? How effective is instruction for students who entered the year with varying levels of performance? How do students in participating schools perform on the third grade MCAS? What are the key factors differentiating students who do and do not attain proficiency on the state s 3 rd grade reading test? 2
Cross-sectional analysis of grade-level changes Changes in the demographic profile over time likely to impact observed outcomes Analysis utilizes a mixed model regression procedure (similar to HLM) controlling for demographic differences in the schools and students being measured Multi-level repeated measures model with observations (students) nested within larger units (schools) Student outcomes (changes over time) modeled as a function of both student-level and school-level factors Statistical significance (p 0.05) indicates that the observed outcome is more than just a function of the change in demography 3
Have student assessment results in participating schools improved over time? Massachusetts relies primarily on results from the DIBELS ORF and GRADE assessments to address the following federal evaluation criteria Increase in percentage of students performing at or above grade-level DIBELS Low Risk and GRADE Average/Strength Decrease in percentage of students with serious reading difficulties DIBELS At Risk and GRADE Weak Overall results show that Massachusetts has met these criteria for all grade-levels 4
DIBELS ORF RF Cohort 1 Percent Low Risk by grade 100 80 60 40 20 50 60 63 65 66 41 50 55 59 60 37 43 47 54 53 0 Grade 1 Grade 2 Grade 3 Spring 2004 Spring 2005 Spring 2006 Spring 2007 Spring 2008 All cumulative changes from 2004 to 2008 are statistically significant 5
DIBELS ORF RF Cohort 1 Percent At Risk by grade 100 80 60 40 20 0 22 36 30 25 23 22 16 14 14 12 28 24 20 17 17 Grade 1 Grade 2 Grade 3 Spring 2004 Spring 2005 Spring 2006 Spring 2007 Spring 2008 All cumulative changes from 2004 to 2008 are statistically significant 6
DIBELS ORF RF Cohort 1 Change in Mean Score (Words Correct per Minute) Spring 2004 Spring 2008 Grade Benchmark N Mean Score N Mean Score Change 1 40 3756 46.43 3688 57.80 11.37 2 90 3679 81.08 3522 95.34 14.26 3 110 3676 97.00 3352 110.20 13.20 All improvements in mean scores from 2004 to 2008 are statistically significant after controlling for demographic shifts over time. All spring 2008 means are higher than spring 2007 means (not shown). In spring 2004 only first grade mean score met the benchmark. By spring 2008, mean scores for all grades are at or above benchmark. 7
DIBELS ORF RF Cohort 2 Percent Low Risk by grade 100 80 60 40 20 48 55 56 60 38 47 53 51 37 42 45 48 0 Grade 1 Grade 2 Grade 3 Spring 2005 Spring 2006 Spring 2007 Spring 2008 All cumulative changes from 2005 to 2008 are statistically significant 8
DIBELS ORF RF Cohort 2 Percent At Risk by grade 100 80 60 40 20 26 21 20 20 39 33 28 30 29 26 24 21 0 Grade 1 Grade 2 Grade 3 Spring 2005 Spring 2006 Spring 2007 Spring 2008 All cumulative changes from 2005 to 2008 are statistically significant 9
DIBELS ORF RF Cohort 2 Change in Mean Score (Words Correct per Minute) Spring 2005 Spring 2008 Grade Benchmark N Mean Score N Mean Score Change 1 40 1821 42.99 1714 50.82 7.83 2 90 1769 77.82 1785 87.17 9.35 3 110 1875 95.55 1746 103.29 7.74 All improvements in mean scores from 2005 to 2008 are statistically significant after controlling for demographic shifts over time. First grade mean scores for both spring 2005 and spring 2008 exceed the benchmark. 10
GRADE Total Test RF Cohort 1 Percent Average/Strength by grade 100 80 60 40 20 62 68 70 71 72 60 63 67 66 69 61 63 66 67 68 0 Grade 1 Grade 2 Grade 3 Spring 2004 Spring 2005 Spring 2006 Spring 2007 Spring 2008 All cumulative changes from 2004 to 2008 are statistically significant 11
GRADE Total Test RF Cohort 1 Percent Weak by grade 100 80 60 40 20 0 25 20 19 18 17 23 20 17 18 15 23 22 18 19 18 Grade 1 Grade 2 Grade 3 Spring 2004 Spring 2005 Spring 2006 Spring 2007 Spring 2008 All cumulative changes from 2004 to 2008 are statistically significant 12
Interpretation of Changes in Mean Standard Score Source: Journal of School Improvement, formerly published by the North Central Association Commission on Accreditation and School Improvement Magnitude of Gain Interpretation 0.10-0.19 SU meaningful; worth mentioning 0.20 0.29 SU quite good 0.30 SU or greater substantial; impressive (www.ncacasi.org/jsi/2000v1i2/standard_score) 13
GRADE Total Test RF Cohort 1 Change in Mean Std. Score Grade N Spring 2004 Spring 2008 Mean Std. Score N Mean Std. Score Change in Std Units Interpretation 1 3729 100.81 3713 104.88 0.27 Quite Good 2 3636 99.08 3536 102.33 0.22 Quite Good 3 3648 99.18 3369 101.43 0.15 Meaningful Standard score of 100 is average for student s grade. Standard deviation of standard score is 15. All changes in mean score (not shown) are statistically significant Interpretation taken from Journal of School Improvement 14
GRADE Total Test RF Cohort 2 Percent Average/Strength by grade 100 80 60 40 20 55 62 62 63 50 54 58 56 49 53 55 55 0 Grade 1 Grade 2 Grade 3 Spring 2005 Spring 2006 Spring 2007 Spring 2008 Cumulative changes for grades 1 and 2 are statistically significant 15
GRADE Total Test RF Cohort 2 Percent Weak by grade 100 80 60 40 20 31 25 24 24 32 27 23 26 35 30 29 26 0 Grade 1 Grade 2 Grade 3 Spring 2005 Spring 2006 Spring 2007 Spring 2008 Cumulative changes for grades 1 and 3 are statistically significant 16
GRADE Total Test RF Cohort 2 Change in Mean Std. Score Grade N Spring 2005 Spring 2008 Mean Std. Score N Mean Std. Score Change in Std Units Interpretation 1 1836 97.50 1702 100.66 0.21 Quite Good 2 1806 94.83 1766 97.56 0.18 Meaningful 3 1918 94.33 1741 97.28 0.25 Quite Good Standard score of 100 is average for student s grade. Standard deviation of standard score is 15. All changes in mean score (not shown) are statistically significant Interpretation taken from Journal of School Improvement 17
GRADE Schools with 80% or more at benchmark All Reading First cohorts Haverhill/Walnut Square (92%) Plymouth/West (90%) Westfield/Moseley (89%) Narragansett/Baldwinville (86%) Plymouth/South (86%) Revere Garfield (85%) Taunton/Walker (84%) Cambridge/Haggerty (82%) Community Day Charter (81%) Methuen/Tenney (81%) Westfield/Franklin Ave (80%) Boston Renaissance (80%) Since they began program implementation, about 70 percent of RF schools have demonstrated increases in the proportion of students in the average/strength category AND decreases in the proportion of students in the weak category. These included about 27 percent of schools which showed substantial improvement, with average/strength increases AND weak decreased of at least 10 percentage points. 18
Is there evidence that RF is closing the performance gap for key demographic subgroups? Nearly demographic subgroups have shown improvement in overall reading skills as measured by GRADE. The exception is for African American students in RF Cohort 2 who have shown a very small decline in A/S performance Of particular note are subgroups with levels of improvement which meaningfully exceed the general population An indication that the performance gap for these students is narrowing Cohort 1: SPED, LEP, Hispanic Cohort 2: LEP There were no subgroups with levels of improvement that were meaningfully smaller than the general population An indication that the performance gap for these students is widening 19
GRADE Total Test Third Grade Subgroups RF Cohort 1 Percent Average/Strength 100 80 60 40 20 23 23 29 28 29 28 26 29 30 34 50 55 58 60 62 0 SPED LEP Low Inc Spring 2004 Spring 2005 Spring 2006 Spring 2007 Spring 2008 Cumulative change for low income students is statistically significant 20
GRADE Total Test Third Grade Subgroups (cont) RF Cohort 1 Percent Average/Strength 100 80 60 40 74 75 78 81 81 57 65 66 61 62 41 45 50 52 57 20 0 White Black Hispanic Spring 2004 Spring 2005 Spring 2006 Spring 2007 Spring 2008 Cumulative change for Hispanic students is statistically significant 21
RF Cohort 1 Subgroups Change in GRADE Mean Std Score 2004 vs. 2008 Group Grade 1 Grade 2 Grade 3 All Students 0.27 0.22 0.15 SPED * 0.46 * 0.33 0.23 LEP * 0.44 0.30 0.08 Low Income 0.35 0.24 0.22 Black 0.26 0.27 0.11 Hispanic * 0.39 0.28 * 0.31 Subgroup results compared to All Students ** Quite good improvement * Meaningful improvement ^ Meaningful lag 22
GRADE Total Test Third Grade Subgroups RF Cohort 2 Percent Average/Strength 100 80 60 40 20 21 29 33 33 44 47 50 52 0 LEP Low Inc Spring 2005 Spring 2006 Spring 2007 Spring 2008 Cumulative changes are not statistically significant 23
GRADE Total Test Third Grade Subgroups (cont) RF Cohort 2 Percent Average/Strength 100 80 60 72 75 75 77 40 37 44 48 49 20 0 White Hispanic Spring 2005 Spring 2006 Spring 2007 Spring 2008 Cumulative change for Hispanic students is statistically significant 24
RF Cohort 2 Subgroups Change in GRADE Mean Std Score 2005 vs. 2008 Group Grade 1 Grade 2 Grade 3 All Students 0.21 0.18 0.25 SPED Insufficient numbers for analysis LEP * 0.33 0.25 0.32 Low Income 0.25 0.24 0.23 Black Insufficient numbers for analysis Hispanic 0.23 0.26 0.29 Subgroup results compared to All Students ** Quite good improvement * Meaningful improvement ^ Meaningful lag 25
How effective is instruction for students who entered the year with varying levels of performance? Effectiveness for Average/Strength Students: calculated for students scoring in the average/strength categories in the fall and provides the percentage of those students who are still scoring at that level in the spring. Effectiveness for Low Average Students: calculated for those students scoring in the low average category in the fall and provides the percentage of those students scoring at the average/strength level in the spring. Effectiveness for Weak Students: is calculated for those students scoring in the weak category in the fall and provides the percentage of those students scoring at low average or above in the spring. Developed by Florida Center for Reading Research (FCRR) using DIBELS. Massachusetts uses GRADE to provide a measure of overall reading ability. 26
Findings: Instructional Effectiveness Among students who began the school year: In Average/Strength categories (stanines 5-9) About 95% ended the year at or above benchmark More than half improved their performance by one or more stanine In the Low Average category (stanine 4) About 70% ended the year in average/strength Instruction had a substantial impact at all grade levels and was most effective for first graders, especially in regard to moving from low average to strength In the Weak Category (stanines 1-3) More than half ended the year in low average or higher Instruction was most effective for first graders, about 47% moved from weak to average/strength 27
Effectiveness for Average/Strength Students (2007-2008 All RF Cohorts) 100 80 60 40 20 0 24 23 46 49 52 68 22 29 93 96 97 19 34 36 15 10 16 9 Grade 1 Grade 2 Grade 3 Declined w/in A/S Same stanine Improved in level Avg to Strength 28
Effectiveness for Low Average Students (2007-2008 All RF Cohorts) 100 80 60 40 20 0 76 6 5 36 66 63 68 63 40 13 25 26 Grade 1 Grade 2 Grade 3 Same stanine Low Avg to Avg Low Avg to Strength 29
Effectiveness for Weak Students (2007-2008 All RF Cohorts) 100 80 60 40 65 13 34 56 25 29 2 44 17 26 1 20 0 18 11 26 21 Grade 1 Grade 2 Grade 3 Improved w/in Weak Weak to Low Avg Weak to Avg Weak to Strength 30
How do students in participating schools perform on the third grade MCAS? Despite improvement on the DIBELS and GRADE, skills have not improved enough to yield improvement on the more challenging MCAS Reading test. Overall performance levels are lower, but the performance trend for RF students is consistent with statewide data showing declines in proficiency and increases in warning. Needs Improvement is more consistent with grade-level performance on nationally-normed assessments In 2008, 89 percent of students statewide met or exceeded NI as did 77 percent of RF students 31
Key Differences: GRADE Level 3 Form B compared to 2008 Grade 3 MCAS Reading Nature of the items GRADE measures a combination of decoding and comprehension skills whereas MCAS is almost exclusively comprehension questions GRADE includes only multiple choice items whereas MCAS also includes two open-response items. Passage difficulty GRADE totals 849 words with an average of 121 words per passage. Passages range from 45 to 196 words. Predominantly text constructed specifically for the test. MCAS totals 4,221 words with an average of 603 words per passage. Passages range from 136 to 1,005 words. All text taken from literature. 32
MCAS Third Grade Reading Test Statewide Results 2003 to 2008 100 80 60 63 63 62 58 59 56 40 20 30 30 31 34 32 33 0 7 7 7 8 9 11 Warning * Needs Improvement* Proficient* 2003 2004 2005 2006 2007 2008 33
MCAS Third Grade Reading Test Cohort 1 Results 2003 to 2008 100 80 60 40 20 0 42 43 42 47 43 43 43 44 44 38 40 36 15 13 14 15 17 21 Warning * Needs Improvement Proficient 2003 2004 2005 2006 2007 2008 34
MCAS Third Grade Reading Test Cohort 2 Results 2004 to 2008 100 80 60 40 20 18 21 22 24 29 48 48 49 47 47 34 31 29 29 24 0 Warning * Needs Improvement Proficient 2004 2005 2006 2007 2008 35
Needs Improvement is more consistent with gradelevel performance on nationally-normed tests GRADE Stanine 2008 MCAS Performance Level (All RF Cohorts) Warning Needs Improvement Proficient or Above 5 9.6% 74.9% 15.5% 6 1.6% 53.0% 45.4% 7 0.0% 24.6% 75.4% 8 0.0% 8.3% 91.7% 9 0.0% 2.2% 97.8% 36
2008 MCAS results School-level Wide disparities in MCAS performance among schools Proficiency 6 schools equal or better than the statewide rate of 56% 28 schools at 25% proficiency or less Warning 11 schools had warning rates equal or better than the statewide rate of 11%, including 3 schools at 0% 19 schools had warning rates of 33% or more Only 11 schools showed both increases in proficiency and decreases in warning 3 schools with substantial improvement (10 or more points) 37
2008 MCAS Top Performing RF Schools Proficiency at or above state average Westfield Moseley (78%) Plymouth South Elem. (75%) Westfield Franklin Ave (73%) Boston Renaissance Charter (65%) North Adams Brayton (60%) Plymouth West Elem. (56%) Statewide proficiency is 56% Warning at or below state average Westfield Moseley (0%) Plymouth South Elem. (0%) Westfield Franklin Ave (0%) Boston Renaissance Charter (3%) Gill-Montague Sheffield (3%) Boston Perkins (3%) Plymouth West Elem. (5%) Chicopee Stefanik (5%) Robert M. Hughes Academy (9%) North Adams Brayton (10%) West Springfield Coburn (11%) Statewide warning is 11% 38
MCAS Schools with Substantial Improvement Proficiency Increases and Warning Decreases of 10+ points School Cohort Proficiency Warning Chicopee Stefanik 1 + 24 pts -17 pts Westfield Moseley 1 + 20 pts -10 pts Lawrence Arlington 1 + 11 pts -15 pts 39
What key factors differentiate students who do and do not attain proficiency on the MCAS? Conducted analysis for all RF and Silber 3 rd graders with spring 2008 GRADE results in the average/strength categories (stanine 5-9) Compared performance of proficient and not-proficient students on the following items: DIBELS ORF: percent low risk GRADE subtests: percent at or above benchmark Individual MCAS passages and test questions (including multiplechoice vs. open-response items) 40
Key Factors in Proficiency All RF and Silber Cohorts GRADE stanine 7-9 GRADE stanine 6 GRADE stanine 5 NP P NP P NP P (333) (1746) (850) (719) (1445) (253) DIBELS ORF percent low risk GRADE Passage Comp percent A/S GRADE Listening Comp percent A/S * 71% 86% * 56% 74% * 46% 59% * 92% 99% *83% 96% * 70% 78% * 70% 83% 66% 70% * 53% 64% * Difference in percentage between proficient and non-proficient students with similar GRADE performance are statistically significant (chi-square) 41
Key Factors in Proficiency (continued) Individual MCAS Passages GRADE stanine 7-9 GRADE stanine 6 GRADE stanine 5 NP P NP P NP P Mean Percent Correct (333) (1746) (850) (719) (1445) (253) MCAS Total Test 69% 84% 64% 80% 59% 79% Passage 4: Star Pictures and Canis Major (poetry) Passage 6: Mercury and the Workmen (play) * 71% 90% 67% 84% 61% 83% 74% 91% * 66% 87% 61% 82% Passage 7: Soil Circle * 56% 82% * 49% 71% * 42% 72% * Difference between NP and P students is disproportionate to the difference in their overall MCAS results (4+ pts greater than the total test mean pct correct) 42
The Open Response Challenge On the two passages with both multiple choice and open response items, RF students perform much better on the multiple choice items than the open response items regardless of their MCAS proficiency and GRADE scores. GRADE stanine 7-9 GRADE stanine 6 GRADE stanine 5 Mean Pct Correct NP P NP P NP P Joanna Cole MC 75% 89% 71% 86% 66% 83% Joanna Cole OR 32% 46% 34% 44% 32% 47% Hello, Goodbye MC 88% 96% 83% 94% 77% 93% Hello, Goodbye OR 42% 56% 41% 55% 40% 60% 43
Findings Opportunities for improving MCAS performance Developing faster and more accurate decoding skills Practicing with longer and more difficult authentic text including high quality expository text Building receptive vocabulary Developing strategies to infer meaning from text Helping students respond to literature especially in writing 44
Summary In Massachusetts, RF has had positive measurable impacts on student skills including improving the performance of students who begin the year at moderate or substantial risk. Yet, it remains important for the state to develop a better understanding of the challenges that limit improvement, particularly on MCAS and provide the necessary PD and support to move forward. Survey responses indicate that RF staff are generally quite positive about the program s impact on their knowledge and practice with regard to effective reading instruction. In the long run, this holds the potential to positively impact students reading skills once program funding is gone. 45
For additional information, please contact: Jennifer Gordon, Research Manager 508-856-1349 jgordon@donahue.umassp.edu UMass Donahue Institute 333 South Street, Suite 400 Shrewsbury, MA 01545 www.donahue.umassp.edu 46