Table 3

Effects on Fountas and Pinnell Benchmark Assessment System Academic Outcomes

 General Text Treatment
(1)		Personalized Text Treatment
(2)		General Text Treatment
(3)		Personalized Text Treatment
(4)
Panel A: Reading Level Outcomes
Reading level (ordinal logit)−0.0638
(0.1478)		0.2753
(0.1811)		0.0133
(0.1769)		0.4915***
(0.1331)
Reading level (standardized point scale)0.0031
(0.0804)		0.1851+
(0.0948)		0.0058
(0.0779)		0.1828*
(0.0782)
Pr(Reading Level A or above)−0.0037
(0.0253)		0.0264
(0.0337)		0.0013
(0.0236)		0.0177
(0.0314)
Pr(Reading Level C or above)0.0115
(0.0417)		−0.0035
(0.0494)		0.0162
(0.0383)		−0.0086
(0.0408)
Pr(Reading Level E or above)−0.0027
(0.0454)		0.0890+
(0.0467)		0.0085
(0.0453)		0.0884*
(0.0385)
Pr(Reading Level G or above)−0.0061
(0.0437)		0.0413
(0.0414)		−0.0101
(0.0444)		0.0414
(0.0388)
Panel B: District Academic Benchmarks
Exceeds expectations−0.0127
(0.0466)		0.1180**
(0.0380)		−0.013
(0.0449)		0.1205***
(0.0331)
Meets or exceeds expectations−0.0052
(0.0411)		0.0869*
(0.0368)		0.0037
(0.0373)		0.0900**
(0.0301)
Approaches, meets, or exceeds expectations0.0263
(0.0508)		0.0094
(0.0466)		0.0291
(0.0474)		0.0074
(0.0410)
Randomization site fixed effects✓✓✓✓
Language of texts✓✓✓✓
Baseline survey controls  ✓✓
Administrative covariates  ✓✓

  • Notes: Each pair of cells represents the results of a separate regression of the treatment effect on the relevant academic outcome. Column headers indicate the model components. Row headers indicate the academic outcome. A graded response model was used to create the factors of baseline survey responses. Factors were made from parent reports of parent age and education; parent reports of the child’s knowledge of letters, letter sounds, and rhyming; parents reports of the frequency with which the parent read to, told stories to, and sang to their child; and parent reports of how often the child asks questions. N = 578 for all regressions. Source data are district test files of the Fountas and Pinnell Benchmark Assessment System in fall of first grade. Standard errors are clustered at the randomization site level.

  • ↵+ p < 0.10,

  • ↵* p < 0.05,

  • ↵** p < 0.01,

  • ↵*** p < 0.001.