## Abstract

Text-message-based parenting programs have proven successful in improving parent engagement and preschoolers’ literacy development. This study seeks to identify mechanisms of the overall effect of such programs. It investigates whether actionable advice alone drives previous studies’ results and whether additional texts of actionable advice improve program effectiveness. The findings provide evidence that text messaging programs can supply too little or too much information. A single text per week is not as effective at improving parenting practices as a set of three texts that also include information and encouragement, but a set of five texts with additional actionable advice is also not as effective as the three-text approach. The results on children's literacy development depend on the child's pre-intervention literacy skills. For children in the lowest quarter of the pretreatment literacy assessments, providing one example of an activity improves literacy scores by 0.19 standard deviations less than providing three texts. Literacy scores of children in higher quarters are marginally higher with only one tip per week than with three tips per week. We find no positive effects of increasing to five texts per week.

## 1.  Introduction

Parents almost invariably aim for their children to succeed in school and beyond, and often are their children's first teacher (Stevenson, Chen, and Uttal 1990). Yet, many parents struggle to provide the necessary support because of limited resources, lack of information, and behavioral challenges. As a result, early home learning environments of children differ substantially (Bradley et al. 2001).1 These differences perpetuate discrepancies in educational attainment and professional success later in life (Heckman 2006). To close learning gaps, a variety of programs has aimed at improving parenting practices. However, many of these parenting programs have shown only limited success, at least in part, due to high demands on parents’ time, infrequency, and information delivery that is difficult for parents to operationalize. Some of the more successful programs are costly and difficult to scale (Karoly et al. 1998; Aos et al. 2004; Duncan, Ludwig, and Magnuson 2010).

Text-messaging interventions have emerged as a promising alternative or supplement due to their low cost, the widespread use of mobile phones, and their ease of scalability. These interventions have been shown to positively influence both student and parent outcomes in a wide array of educational settings.2 In particular, a text-messaging program developed at Stanford University improved parent engagement and children's literacy development in preschool (Doss et al. 2019; York, Loeb, and Doss 2019). The program breaks down the complexities of parenting and thereby overcomes informational and behavioral barriers that may inhibit parents from effective parenting.3 It provides a combination of general information about important literacy skills and parent-child activities with FACT text messages, actionable advice with specific examples of parent-child literacy activities with TIP text messages, and encouragement/reinforcement with GROWTH text messages.

The original program, first evaluated by York, Loeb, and Doss (2019) in the San Francisco Unified School District, sent a FACT message on Mondays, a TIP message on Wednesdays, and a GROWTH message on Fridays to parents of pre-school students over eight months. We will call this program the FACT-TIP-GROWTH (FTG) program (or original program). The authors found substantial positive effects of the program on children's early literacy skills. These positive effects were largely driven by children who started the year in the lower half of literacy development. Doss et al. (2019) tested whether the content of texts mattered or whether the benefits were driven solely by the reminder of getting a text about parenting. The study compared the original FTG program to one that provided tips that better matched the skill level of the children. The authors found that targeting texts based on skills improved results even further, as children who started the year in the bottom or top third of the literacy distribution benefited more from the differentiated program, whereas those who started in the middle third did not.

We extend prior findings by answering two questions focusing on the importance of content and frequency of the text messages. First, does the actionable advice (i.e., examples of activities) in the TIP message drive previous results or is the addition of general information and encouragement/reinforcement through the FACT and GROWTH texts more important for increasing parent–child interactions and child development? Second, does the provision of more activities through two additional TIP messages further improve parent–child interactions and child development? Answering these questions will shed light on how to best overcome behavioral barriers that parents face in providing strong home learning environments for their children and can guide a more efficient and effective program design. It will further uncover the process of parent behavior change that can inform a much broader array of programs. Moreover, this paper broadly speaks to the question of whether schools can inundate parents with too much information.

We study these two interrelated questions in a randomized experiment comparing two new variations of the program to the FTG program. We assign parents of prekindergarten children into three experimental groups. We use a blocked randomization design and assign a third of parents within each preschool into each treatment group. The first group of parents only receives one TIP message on Wednesdays, henceforth the Tip program. The second group receives the FTG program (i.e., the original program). The third group receives the FACT message on Mondays, TIP messages on Tuesdays, Wednesdays, and Thursdays, and the GROWTH message on Fridays (i.e., the original program and two additional TIP messages), henceforth the FACT-TIP-TIP-TIP-GROWTH (FTTTG) program. Table 1 provides examples of text messages for each treatment group that focuses on letter recognition.

Table 1.

Examples of Text Messages for Each Treatment Group

Text Messages by Treatment Program:
Week DayTip ProgramFTG Program (original program)FTTTG Program
Monday  FACT: Children need to know letters to learn how to read & write. Research shows kids with good letter knowledge become good readers. FACT: Children need to know letters to learn how to read & write. Research shows kids with good letter knowledge become good readers.
Tuesday   TIP: Point out the first letter in your child's name in magazines, on signs & at the store. Have your child try. Make it a game. Who can find the most?
Wednesday TIP: Point out the first letter in your child's name in magazines, on signs & at the store. Have your child try. Make it a game. Who can find the most? TIP: Point out the first letter in your child's name in magazines, on signs & at the store. Have your child try. Make it a game. Who can find the most? TIP: Now point out each of the letters in your child's name. After you point to a letter ask: What sound does it make?
Thursday   TIP: See if your child can name the letters on common objects like a stop sign. Can s/he name all of the letters on the sign with your street's name?
Friday  GROWTH: Keep pointing out letters. You're preparing your child 4K! Point out each of the letters in your child's name. Ask: What sound does it make? GROWTH: Keep pointing out letters to prepare your child 4K! Now have him/her make the letter sounds (ss, tt, oo & pp). What other words have those sounds?
Text Messages by Treatment Program:
Week DayTip ProgramFTG Program (original program)FTTTG Program
Monday  FACT: Children need to know letters to learn how to read & write. Research shows kids with good letter knowledge become good readers. FACT: Children need to know letters to learn how to read & write. Research shows kids with good letter knowledge become good readers.
Tuesday   TIP: Point out the first letter in your child's name in magazines, on signs & at the store. Have your child try. Make it a game. Who can find the most?
Wednesday TIP: Point out the first letter in your child's name in magazines, on signs & at the store. Have your child try. Make it a game. Who can find the most? TIP: Point out the first letter in your child's name in magazines, on signs & at the store. Have your child try. Make it a game. Who can find the most? TIP: Now point out each of the letters in your child's name. After you point to a letter ask: What sound does it make?
Thursday   TIP: See if your child can name the letters on common objects like a stop sign. Can s/he name all of the letters on the sign with your street's name?
Friday  GROWTH: Keep pointing out letters. You're preparing your child 4K! Point out each of the letters in your child's name. Ask: What sound does it make? GROWTH: Keep pointing out letters to prepare your child 4K! Now have him/her make the letter sounds (ss, tt, oo & pp). What other words have those sounds?

Notes: To ensure comparability of treatments, the text message content is similar across treatment groups. The Monday FACT texts are the same in both the original and the FACT-TIP-TIP-TIP-GROWTH (FTTTG) programs. The Tip program and the original program send the same TIP messages on Wednesdays. However, rearrangement and adaptation of a few texts are necessary in the FTTTG program in order to achieve a sensible progression of activities. Specifically, the FTTTG program sends Wednesday's TIP messages on Tuesdays, and the example in the GROWTH message is used as Wednesday's TIP message. More text examples can be found in York, Loeb, and Doss (2019) and in Doss et al. (2019). FTG = FACT-TIP-GROWTH.

We ran this study in partnership with the Dallas Independent School District (hereafter referred to as Dallas ISD). Parents of 4-year-old preschoolers in the district received the texting intervention during the 2015–16 school year. The intervention launched in mid-November 2015 and delivered text messages through 24 June 2016.4 Parents were able to choose to receive texts in English or Spanish.

The text messages cover a wide range of literacy skills and related parenting practices, including: uppercase and lowercase letter recognition, letter sound awareness, beginning sound awareness, rhyme, name writing, concepts of print, story comprehension, vocabulary, singing and listening to songs, self-narration, parent–child conversations, and parent–child book reading routines. These skills and activities align with Texas state standards for early literacy skills (e.g., Texas Education Agency 2015).5 The texting curriculum is structured as a spiral curriculum—it starts simple and becomes progressively more advanced over the eight months of the intervention—and topics are reintroduced throughout the year for reinforcement. Most of the texts relate to existing family routines and activities (e.g., bath time, commuting and travel, or family meals) to minimize costs of adopting beneficial behavior.6

We find that the original FTG program has benefits in comparison to the Tip program, suggesting that the other elements of the FTG program are helping parents. Results from a parent survey suggest that providing only one TIP leads to lower self-reported parent engagement than in the FTG program. This finding provides evidence that providing context and encouragement in addition to advice may maintain interest and engagement by highlighting the importance of the respective skill and by providing positive reinforcement. Parents who only receive advice may not use it because they might not understand its relevance. However, the additional TIPs in the FTTTG program are also not beneficial for parents and, in some cases, may be detrimental. The negative effects of additional texts are seen in the opt out rates of parents. In comparison with the FTG program, significantly more parents assigned to the FTTTG program discontinued receiving the texts, whereas significantly fewer parents assigned to the Tip program did so.

The effects on child literacy development depend on the child's pre-intervention literacy skills. For children in the lowest quarter of the pretreatment literacy assessments, the Tip program improves literacy scores by 0.19 standard deviations (SD) less than also providing general information, encouragement, and reinforcement in the FTG program. These results demonstrate that context and encouragement may be important to parents of lower-performing children. The literacy scores of children in the middle two quarters are higher in the Tip program than in the original program, but these effects are only marginally significant. We find no effects for providing additional examples of activities in the FTTTG program on children's literacy test scores.

## 2.  Data and Descriptive Statistics

The Dallas ISD is the second largest public school district in Texas, and the fourteenth largest district in the nation.7 The district serves approximately 10,000 prekindergarten students in one hundred thirty-two preschools. The prekindergarten student population is diverse and economically disadvantaged. The main eligibility criteria for prekindergarten enrollment are that children are unable to speak and comprehend the English language or that children are eligible to participate in the National School Lunch Program.

### Data Sources

In this study, we use information about the children, their parents, and their teachers. Overall, 3,473 families were part of our study. Parent information comes from three main sources. First, we obtained preferred texting language, age, and highest educational attainment from our enrollment forms. Second, we gathered the opt-out information from our texting platform, EZtexting.com. Parents were able to opt out by replying “Stop” or similar words to any text message. We use this opt-out information as a measure of parents’ overall experience of the texting program. Third, we surveyed parents after the texting intervention ended at the end of the school year. We collected measures of parent–child engagement, such as reading and literacy activities, and overall satisfaction with the texting program. Parents were invited to participate in the survey by text, e-mail, and regular mail during the months of August through September 2016. We offered parents \$20 for completing the survey. Ultimately, 664 parents did so. For our analysis, we only consider the 648 parents who answered all questions, a response rate of 18.6 percent. Though the survey response rate is low, the treatment status did not affect survey participation (see Attrition Analysis in section 4). Parents who did not answer the survey are, on average, less educated and older, and they are less likely to be black and more likely to be Hispanic.8 Hence, the results of the parent survey outcomes may not be fully representative of the overall sample.

The parent survey included four series of questions. The first addressed the parents’ experience with the texting program. For instance, the survey asked if parents received and read the text messages, and if parents used the information and found it helpful. It also asked if parents would have liked to have received more or less information. The second series of questions asked about parents’ confidence in supporting their child's school readiness in literacy and math skills, and in improving their child's behavior.9 The third series of questions assessed the frequency of activities when reading a book to their child, such as letting the child turn the pages, talking about pictures, asking questions, and underlining words with the finger. The fourth series of questions assessed the frequency of reading related activities, including among others helping the child to write her name, practicing word sounds, and helping to learn more words.

The child information comes from the Dallas ISD administrative student records. These data include demographic information, such as age, gender, race and ethnicity, and an indicator for low socioeconomic status.10 The data also include our main child literacy outcome measure, the Circle Assessment System (hereafter referred to as Circle). Circle is a one-on-one literacy assessment that takes less than 10 minutes per child to complete. According to the Dallas ISD staff, the Circle assessments are used to monitor student achievement at the school, teacher, and student levels. Although there are no sanctions or rewards attached to performance on the assessment in an accountability framework, Circle scores are shared with teachers, school administrators, and Early Learning Department to inform instructional support. All children in this study were assessed with either the English- or Spanish-language version of Circle.11 Circle assesses language and literacy skills along three distinct dimensions: (1) rapid letter naming, (2) rapid vocabulary naming, and (3) phonological awareness. Specifically, the rapid letter naming task measures a child's alphabet knowledge (a 1-minute timed assessment task); the rapid vocabulary naming task evaluates a child's ability to name common objects (a 1-minute timed assessment task); and the phonological awareness task assesses a child's understanding of sound (approximately 5 minutes).12 The phonological awareness is a sum of the following four subtasks:13rhyming (i.e., the ability to distinguish if two words rhyme when spoken), alliteration (i.e., the ability to indicate same beginning sound(s) between two or more words), syllabication (i.e., the ability to separate a word into parts), and onset-rime (i.e., the ability to blend two parts of a word together when segmented between the beginning consonant[s] and the rest of the word). For Spanish speakers, the phonological awareness assessment only includes rhyming, alliteration, and syllabication.

The Circle assessment is administered three times during the school year: Circle-1 is carried out in the beginning of the year (September/October 2015), Circle-2 is carried out in the middle of the year (January/February 2016), and Circle-3 is carried out at the end of the year (April/May 2016). Because the first assessment of Circle occurred before the intervention started, we use Circle-1 results as covariates in all regression specifications. Our main set of child outcomes comes from the third assessment (Circle-3), as parents and children had the most exposure to the program.14,15

The teacher information also comes from the Dallas ISD administrative data. These data include teachers’ gender, race/ethnicity, experience in years, and the number of hours they were absent in the school year. For each child, we use mean characteristics of all of their teachers during the school year.

### Descriptive Statistics

Table 2 presents descriptive statistics on the sample of parents represented in the randomization sample (N = 3,473), Circle-3 test sample (N = 2,920), and parent survey sample (N = 648) by children, parent, and teacher characteristics. As shown in the first column of table 2, about 11 percent of children in the randomization sample are black, 85 percent are Hispanic, and 2 percent are Asian and white, respectively, and the majority of the sample is of low socioeconomic status (95 percent). The average fall age of children in this sample is 4.7 years.

Table 2.

Summary Statistics—–Means and Standard Deviations

Randomization SampleCircle-3 Test SampleParent Survey Sample
Panel A: Children Characteristics
Age (in years) 4.68 4.71 4.68
(0.33) (0.29) (0.34)
Female 0.50 0.50 0.48
Race and ethnicity
Black 0.11 0.10 0.13
Hispanic 0.85 0.86 0.83
Asian 0.02 0.02 0.01
White 0.02 0.02 0.02
Other 0.01 0.00 0.01
Low socioeconomic status 0.95 0.95 0.95
Circle-1 assessment (raw) scoresa
Rapid letter naming (RLN) 4.77 4.73 4.97
(7.33) (7.50) (7.49)
Missing RLN 0.07 0.00 0.07
Rapid vocabulary (RV) 10.33 10.27 10.82
(6.74) (7.00) (6.98)
Missing RV 0.07 0.00 0.08
Phonetic awareness (PA) 10.13 10.22 10.27
(4.86) (5.08) (4.96)
Missing PA 0.08 0.00 0.09
Panel B: Parent Characteristics
Age (in years)a 31.16 31.27 30.69
(5.89) (5.83) (5.96)
Missing age 0.23 0.24 0.18
Highest education levelb
Less than high school 0.30 0.31 0.28
High school 0.26 0.26 0.39
Some college 0.13 0.12 0.19
Associate degree 0.04 0.04 0.06
Bachelor's degree 0.03 0.03 0.05
Missing education 0.22 0.23 0.01
Texting language
English 0.36 0.34 0.38
Spanish 0.64 0.66 0.62
Panel C: Teacher Characteristics
Average female 0.76 0.76 0.78
Average race and ethnicity
Black 0.19 0.19 0.20
Hispanic 0.46 0.46 0.44
White 0.32 0.32 0.33
Other 0.03 0.03 0.07
Average missing teacher demographics 0.01 0.01 0.01
Average experience (in years) 8.36 8.49 8.27
(5.88) (5.96) (5.79)
Average hours of absence useda 49.64 49.86 48.90
(33.27) (34.48) (33.37)
Average missing hours of absence 0.12 0.12 0.14
Observations 3,473 2,920 648
Randomization SampleCircle-3 Test SampleParent Survey Sample
Panel A: Children Characteristics
Age (in years) 4.68 4.71 4.68
(0.33) (0.29) (0.34)
Female 0.50 0.50 0.48
Race and ethnicity
Black 0.11 0.10 0.13
Hispanic 0.85 0.86 0.83
Asian 0.02 0.02 0.01
White 0.02 0.02 0.02
Other 0.01 0.00 0.01
Low socioeconomic status 0.95 0.95 0.95
Circle-1 assessment (raw) scoresa
Rapid letter naming (RLN) 4.77 4.73 4.97
(7.33) (7.50) (7.49)
Missing RLN 0.07 0.00 0.07
Rapid vocabulary (RV) 10.33 10.27 10.82
(6.74) (7.00) (6.98)
Missing RV 0.07 0.00 0.08
Phonetic awareness (PA) 10.13 10.22 10.27
(4.86) (5.08) (4.96)
Missing PA 0.08 0.00 0.09
Panel B: Parent Characteristics
Age (in years)a 31.16 31.27 30.69
(5.89) (5.83) (5.96)
Missing age 0.23 0.24 0.18
Highest education levelb
Less than high school 0.30 0.31 0.28
High school 0.26 0.26 0.39
Some college 0.13 0.12 0.19
Associate degree 0.04 0.04 0.06
Bachelor's degree 0.03 0.03 0.05
Missing education 0.22 0.23 0.01
Texting language
English 0.36 0.34 0.38
Spanish 0.64 0.66 0.62
Panel C: Teacher Characteristics
Average female 0.76 0.76 0.78
Average race and ethnicity
Black 0.19 0.19 0.20
Hispanic 0.46 0.46 0.44
White 0.32 0.32 0.33
Other 0.03 0.03 0.07
Average missing teacher demographics 0.01 0.01 0.01
Average experience (in years) 8.36 8.49 8.27
(5.88) (5.96) (5.79)
Average hours of absence useda 49.64 49.86 48.90
(33.27) (34.48) (33.37)
Average missing hours of absence 0.12 0.12 0.14
Observations 3,473 2,920 648

Notes: Numbers in parentheses are standard deviations for continuous variables. Circle-3 test sample is conditioned on having Circle-1 test scores.

aMissing data values imputed to be the mean.

bMissing data values set to zero.

Also shown in table 2 are the child's pre-intervention Circle-1 raw test scores. Children, on average, name 4.8 letters in 1 minute, identify 10.3 vocabulary objects in 1 minute, and get 10.1 items correct in the phonetic awareness assessment. Separating these scores based on the language (English or Spanish) used to assess (not shown), we find: English language–assessed children were able to name 7.23 letters in 1 minute, identify 15.3 vocabulary objects in 1 minute, and get 11.7 items correct in the phonetic awareness assessment; Spanish language–assessed counterparts name 3.4 letters in 1 minute, identify 7.8 vocabulary objects in 1 minute, and get 9.4 items correct in the phonetic awareness. To put these tests into context, according to the Circle's technical manual as of 9 October 2018, an English-speaking child between the ages of 4 and 4.5 years should be able to name 7 letters in 1 minute, identify 16 vocabulary objects in 1 minute, and get 8 items correct in the phonetic awareness assessment at the beginning of the school year. A Spanish-speaking child between the ages of 4 and 4.5 should be able to name 4 letters in 1 minute, identify 6 vocabulary objects in 1 minute, and get 5 items correct in the phonetic awareness assessment.16 Thus, at the beginning of the intervention, English language–assessed children had lower-than-average scores in rapid vocabulary, and Spanish language-assessed children had lower-than-average scores in rapid letter naming.

Regarding parent characteristics (shown in panel B), 30 percent have less than a high school degree, 26 percent have a high school degree, and 22 percent have some college or higher. The average fall age of parents in this sample is 31.2 years. About two thirds of parents, 64 percent, chose to receive texts in Spanish, and 36 percent chose English.

As for average teacher characteristics (shown in panel C), most teachers are female (76 percent), and the average experience is 8.4 years in the district. The racial and ethnic teacher composition in the district differs from that of the student population: 19 percent are black and 46 percent are Hispanic, but a higher percentage of teachers were white (32 percent) compared with the student population. A side-by-side comparison of all three samples by child, parent, and teacher characteristics shows similar characteristics by most of these covariates.

To assess the representativeness of our results, table B.1 in our online appendix shows how different our randomization sample (students whose parents consented to receive the text messages) is from those preschool students whose parents did not consent to participate in the experiment. The most striking differences are that our randomization sample consists of 20 percent more Hispanic students, 18 percent fewer black students, and 5 percent more socioeconomically disadvantaged students. These differences are also reflected in students’ teacher characteristics. Students in our randomization sample have, on average, less often black teachers, more often Hispanic teachers, teachers with more experience, and teachers who are absent less frequently. As a result, our results may not generalize to the entire preschool student body and be more representative of a more Hispanic and slightly less affluent student body.

## 3.  Empirical Strategy

### Estimating Treatment Effects

We estimate the treatment effects of one specific activity in the Tip program and of two additional activities in the FTTTG program in comparison to the original texting program, FTG, with the following model specification:
$yis=α+β1·Tipis+β2·FTTTGis+δ·Xis+γs+ɛis,$
where $yis$ is the outcome of interest of parent (or child) $i$ in preschool site $s$. The main parent outcomes are survey measures of parent–child engagement and program experience as well as parents who opt out of the program. The child outcomes are the Circle-3 literacy test scores, standardized within testing language (English or Spanish) to have standard deviation one and mean zero.17 The variables $Tipis$ and $FTTTGis$ are binary indicators of whether a parent received the Tip or FTTTG programs in comparison to receiving the original program (omitted category in all regression specifications, FTG program), respectively. $Xis$ is a vector of covariates that includes child characteristics (i.e., age, gender, race/ethnicity, low-socioeconomic status, and pre-intervention Circle-1 test scores), parent characteristics (i.e., age and highest educational attainment), and lastly, teacher characteristics (i.e., gender, race/ethnicity, years of experience, and hours absent in school year). $γs$ are preschool site fixed effects and $ɛis$ is a parent-level (or child-level) error-term. Standard errors are clustered at the pre-school site level. The coefficients of interest are $β1$ and $β2$. They measure the causal effects of receiving Tip or FTTTG programs in comparison to original texting intervention, the FTG program. Note that we cannot report treatment effects of the programs in comparison to not receiving a program because we did not have a pure control group. Moreover, given that some parents may not receive the texts or will not pay attention to them, the program effects are intention-to-treat effects.18 In online table B.2, we report on the results of preschool site fixed effects models to evaluate balance of child, parent, and teacher characteristics. Most coefficients are small and insignificant and we cannot reject the null hypothesis that the coefficients are jointly equal to zero providing evidence of successful randomization.19

To explore treatment effect heterogeneity, we also estimate the above model specification in quarters of the child's pre-intervention Circle-1 test. Parents of children who start prekindergarten less prepared than their peers may struggle more with the aforementioned behavioral barriers than other parents. In that case, providing text messages are aimed at overcoming these barriers may be particularly effective for parents of lower-performing children.

### Attrition Analysis

We analyze attrition in both the child outcome data (i.e., Circle-3 test sample) and parent outcome data (i.e., parent survey sample) by testing whether attrition differs by treatment status. For our purposes, attrition simply means that families who are included in the randomization sample are not part of the Circle-3 test sample or the parent survey sample, respectively. See section 2 for details on reasons parents were excluded from the respective samples. Note that parents who opted out of receiving the text messages are still part of the Circle-3 test sample and, because we surveyed parents also via e-mail and mail, are likely part of the parent's survey sample. If the attrition rate of parents who received the Tip or FTTTG programs systematically differ to that of the comparison group (i.e., FTG program) in a way that is related to our study outcomes, then our treatment effects would be biased.

Thus, to check for selective attrition from the Circle-3 test sample and parent survey sample, we estimate the following preschool site fixed effects regression models:
$Ais=α+β1·Tipis+β2·FTTTGis+δ·Xis+γs+ɛis,$
where $Ais$ is a binary indicator that equals one (and zero otherwise) if a child (or parent) does not appear in the Circle-3 test sample or parent survey sample.20 We also control for child, parent, and teacher characteristics, and preschool site fixed effects.

Table 3 reports the results for the attrition analysis. The first row shows that attrition in the Circle-3 test sample overall does not differ by treatment status (panel A). Both coefficients are close to zero and are statistically insignificant. However, when examining attrition by quarters of Circle-1 literacy scores, the coefficients for the Tip treatment are marginally statistically significant for the two middle quarters: 4.4 percentage points in the second quarter and −4.7 percentage points in the third quarter. However, the effect on the attrition pattern does not appear systematic and may simply be due to noise.21 That said, given this differential attrition in these two quarters, we assess the robustness of our main results with a bounding analysis in section 5. Our results remain largely unchanged. Panel B of table 3 shows that neither treatment group is affected by selective attrition into the parent survey. All reported estimates are statistically insignificant.

Table 3.

Treatment Effects on Attrition in Circle-3 Test Sample and Parent Survey Sample

Panel A: Circle-3 Test Sample—–Whether any Circle-3 Test Outcomes are Missing
Full sample −0.004 −0.006 0.15 0.59 3,473
(0.010) (0.009)
By quartiles:a
Quarter 1 (lowest) −0.028 −0.017 0.10 0.21 780
(0.025) (0.027)
Quarter 2 0.044* 0.002 0.04 0.32 778
(0.022) (0.017)
Quarter 3 −0.047* −0.029 0.09 0.12 779
(0.024) (0.026)
Quarter 4 (highest) 0.011 0.023 0.03 0.01 777
(0.019) (0.020)
Panel B: Parent Survey Sample—–Whether any Parent Survey Outcomes are Missing
Entire sample 0.012 0.000 0.81 0.10 3,473
(0.015) (0.016)
Panel A: Circle-3 Test Sample—–Whether any Circle-3 Test Outcomes are Missing
Full sample −0.004 −0.006 0.15 0.59 3,473
(0.010) (0.009)
By quartiles:a
Quarter 1 (lowest) −0.028 −0.017 0.10 0.21 780
(0.025) (0.027)
Quarter 2 0.044* 0.002 0.04 0.32 778
(0.022) (0.017)
Quarter 3 −0.047* −0.029 0.09 0.12 779
(0.024) (0.026)
Quarter 4 (highest) 0.011 0.023 0.03 0.01 777
(0.019) (0.020)
Panel B: Parent Survey Sample—–Whether any Parent Survey Outcomes are Missing
Entire sample 0.012 0.000 0.81 0.10 3,473
(0.015) (0.016)

Notes: Each row represents a separate regression model (only the coefficients of the treatments status are reported). All regressions include controls for student characteristics (age, gender, race/ethnicity, low socioeconomic status, and Circle-1 test scores), parent characteristics (age and highest educational attainment), teacher characteristics (gender, race/ethnicity, years of experience, and hours of absence used), and preschool site fixed effects. Standard errors are clustered at the preschool site level. The omitted reference group in all regressions is the original texting program of three texts per week (i.e., FTG program). FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

aQuarters are based on student's Circle-1 tests prior to the intervention.

*p < 0.10.

## 4.  Main Results

### Results on Parent Program Experience and Engagement

We find evidence that the experience of some parents was tempered by increasing the number of text messages they received per week. As shown in table 4 (panel A), for the Circle-3 sample, parents who received the Tip program were 2 percentage points less likely to opt out of the program compared with those receiving the original program, FTG. Parents who received the FTTTG program were about 2.9 percentage points more likely to opt out compared with those receiving the original program. Furthermore, these results are driven mostly by parents whose children are in the lowest quarter of their pre-intervention literacy assessment. These parents were 4.4 percentage points less likely to opt out if they were in the Tip program, and 4.9 percentage points more likely to opt out of the FTTTG program compared with the FTP program.

Table 4.

Treatment Effects on Parents’ Opting Out of the Intervention and Overall View of Receiving the Weekly Text Messages

Panel A: Parent Opt-Out Rates
Full sample −0.020** 0.029*** 0.05 0.03 2,920
(0.008) (0.010)
By quartiles:a
Quarter 1 (lowest) −0.044** 0.049** 0.04 0.09 715
(0.019) (0.020)
Quarter 2 −0.011 0.01 0.04 −0.04 728
(0.020) (0.024)
Quarter 3 −0.038* 0.003 0.05 −0.01 735
(0.022) (0.026)
Quarter 4 (highest) −0.019 0.032 0.05 −0.01 742
(0.024) (0.026)
Panel B: Parent Survey Responses to Receiving the Weekly Text Messages
Tip FTTTG FTG Mean and SD Adjusted R2 N
Received texts (0/1) −0.022 −0.013 0.98 0.10 648
(0.022) (0.016)
Read texts (SD) −0.188* −0.250** 3.59 −0.04 648
(0.111) (0.119) (0.67)
Uses texts (SD) −0.134 −0.169 3.02 −0.02 648
(0.127) (0.128) (0.78)
Texts are helpful (SD) −0.134 0.070 3.44 0.05 648
(0.122) (0.112) (0.65)
Receive too many texts (0/1) −0.021 0.038 0.06 0.02 648
(0.027) (0.028)
Not enough texts (0/1) 0.041 0.016 0.04 0.03 648
(0.026) (0.025)
Ideal number of texts (SD) −0.547*** −0.158 4.27 0.03 648
(0.120) (0.111) (1.50)
Recommend texts (SD) −0.132 −0.043 3.45 −0.01 648
(0.111) (0.125) (0.54)
SUR test for joint significance (p-value)b 0.000 0.020
Panel A: Parent Opt-Out Rates
Full sample −0.020** 0.029*** 0.05 0.03 2,920
(0.008) (0.010)
By quartiles:a
Quarter 1 (lowest) −0.044** 0.049** 0.04 0.09 715
(0.019) (0.020)
Quarter 2 −0.011 0.01 0.04 −0.04 728
(0.020) (0.024)
Quarter 3 −0.038* 0.003 0.05 −0.01 735
(0.022) (0.026)
Quarter 4 (highest) −0.019 0.032 0.05 −0.01 742
(0.024) (0.026)
Panel B: Parent Survey Responses to Receiving the Weekly Text Messages
Tip FTTTG FTG Mean and SD Adjusted R2 N
Received texts (0/1) −0.022 −0.013 0.98 0.10 648
(0.022) (0.016)
Read texts (SD) −0.188* −0.250** 3.59 −0.04 648
(0.111) (0.119) (0.67)
Uses texts (SD) −0.134 −0.169 3.02 −0.02 648
(0.127) (0.128) (0.78)
Texts are helpful (SD) −0.134 0.070 3.44 0.05 648
(0.122) (0.112) (0.65)
Receive too many texts (0/1) −0.021 0.038 0.06 0.02 648
(0.027) (0.028)
Not enough texts (0/1) 0.041 0.016 0.04 0.03 648
(0.026) (0.025)
Ideal number of texts (SD) −0.547*** −0.158 4.27 0.03 648
(0.120) (0.111) (1.50)
Recommend texts (SD) −0.132 −0.043 3.45 −0.01 648
(0.111) (0.125) (0.54)
SUR test for joint significance (p-value)b 0.000 0.020

Notes: Each row represents a separate regression model (only the coefficients of the treatments status are reported). All regressions include controls for student characteristics (age, gender, race/ethnicity, low socioeconomic status, and Circle-1 test scores), parent characteristics (age and highest educational attainment), teacher characteristics (gender, race/ethnicity, years of experience, and hours of absence used), and preschool site fixed effects. Standard errors are clustered at the preschool site level. The omitted reference group in all regressions is the original texting program of three texts per week (i.e., FTG program). FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

aQuarters are based on student's Circle-1 tests prior to the intervention.

bThe seemingly unrelated regression (SUR) test for joint significance is testing the joint significance of each program on all survey items in table 4. For standardized measures, means and standard deviations (SD) are calculated from the underlying nonstandardized measures.

*p < 0.10; **p < 0.05; ***p < 0.01.

Figure 1 shows graphically the timing of when parents opted out from receiving the text messages in each of the three programs. It displays the cumulative shares of parents who have opted out (y-axis) given the number of days parents have been in the programs (x-axis). The figure highlights that opt-out is similarly low immediately after the start of all three programs. However, over the course of the program, as parents receive different amounts of text messages, the differences in the shares of parents who opt out among the programs gradually widen, particularly so for the FTTTG program. Thus, figure 1 provides further evidence that, over time, more text messages can temper parents’ experience and lead them to no longer participate in the program.

#### Cumulative Parent Opt-out by Program

Figure 1.
Cumulative Parent Opt-out by Program

Notes: FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

Figure 1.
Cumulative Parent Opt-out by Program

Notes: FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

Next, panel B of table 4 shows supporting evidence for the opt-out results based on the parent survey data. In the survey, we asked parents about their overall experience with the text messages.22 For the most part, parent's responses to these questions align with the parent opt-out data. For instance, the FTTTG program decreased the frequency of parents’ reading the texts by 0.25 SD compared with the original texting program. Interestingly, parents who received only the Tip program also decreased the frequency of reading the texts by 0.19 SD compared with the original texting program. These parents also reported that the “ideal” number of texts should be lower, compared with parents in the original program. Overall, although some parents opted out more with three texts than with one text, parents reported that they like three texts per week the best. Five texts per week increased opt out and was not preferred to three texts per week in the survey reports.

We further find evidence of the benefits of three texts relative to one or five when looking at parent-reported interactions with their child. Table 5 shows the treatment effects on parent's self-reported reading and literacy activities, and parent understanding of child development based on the parent survey.23 As shown across all panels of table 5, parents in the Tip program did reading and literacy activities less frequently with their child than parents in the original program. Parents reported lower frequencies for all activities, and significantly so for: talking about the pictures in a book, asking their child questions about what is happening in the story, underlining words with their finger as they read to their child; practicing word sounds with their child (e.g., milk starts with “mmm”), asking their child about their day, and helping their child learn new words. Whereas the effects on average scores for both reading and literacy activities are both negative and sizable, only the effect on the average literacy activity score is statistically significant (–0.28 SD). Parents in the Tip program also indicated a significantly lower average confidence in how to develop their child's skills compared with the original program (–0.22 SD). The FTTTG program decreased self-reported frequency by which parents underline words as they read and read to their child. Although almost all effects on individual items are negative, the effects on average scores are insignificant. A test of whether the effects of each program on all individual survey items are significantly different from zero is rejected for the Tip program but not for the FTTTG program. Altogether, these survey results suggest that providing only one actionable activity leads to lower self-reported parent reading and literacy activities in comparison to the original program, which scaffolds activities with information and encouragement. However, providing additional actionable activities does not improve parent–child engagement and may even reduce engagement.

Table 5.

Treatment Effects on Parent Reading Activities, Literacy Activities, and Understanding of Child Development

Panel A: Frequency of How Often Parents Did Reading Activities with their Child (standardized)
Mean Reading Activity −0.202 −0.156 — 0.06 648
(0.127) (0.113)
Let child hold the book −0.052 −0.160 2.82 0.03 648
(0.107) (0.117) (0.81)
Show child book parts −0.073 −0.142 3.05 0.04 648
(0.113) (0.119) (0.94)
Talk about the pictures in a book −0.218* −0.046 3.39 0.01 648
(0.126) (0.108) (0.75)
(0.119) (0.105) (0.81)
Underline words as you read −0.269** −0.299** 3.30 0.08 648
(0.106) (0.118) (0.83)
Show child that we read from left to right −0.112 −0.115 3.24 0.00 648
(0.127) (0.118) (1.01)
(0.129) (0.103) (0.99)
Panel B: Frequency of How Often Parents Did Literacy Activities with their Child (standardized)
Mean Literacy Activity −0.281** −0.144 — 0.06 648
(0.130) (0.122)
Helped child write his/her name −0.021 −0.041 2.56 −0.04 648
(0.126) (0.131) (0.94)
Pointed out letters on objects −0.195 −0.110 2.91 0.01 648
(0.126) (0.128) (0.92)
Sang a song or nursery rhyme with child −0.107 −0.099 2.87 −0.02 648
(0.126) (0.130) (0.92)
Practiced word sounds −0.309** −0.071 2.95 0.03 648
(0.119) (0.125) (0.92)
Said rhyming words with child −0.188 −0.101 2.77 0.04 648
(0.120) (0.123) (0.97)
Asked child about his or her day −0.357*** −0.203 3.45 0.03 648
(0.122) (0.123) (0.76)
Played a literacy learning game with child −0.148 −0.041 2.52 0.04 648
(0.118) (0.108) (1.01)
Helped child learn new words −0.245* −0.023 3.02 0.07 648
(0.126) (0.119) (0.90)
Worked on reading skills during family activities −0.293** −0.190 2.78 0.10 648
(0.118) (0.120) (0.95)
Helped child sound out a word −0.147 −0.074 2.91 0.05 648
(0.130) (0.122) (0.94)
Read to child −0.211* −0.188* 3.08 0.04 648
(0.119) (0.110) (0.87)
Panel C: Parents’ Understanding of How to Develop Skills for their Child (standardized)
Mean Skill Development −0.219* 0.034 — 0.11 648
(0.115) (0.101)
Literacy development −0.210* 0.009 3.28 0.06 648
(0.112) (0.115) (0.71)
Math development −0.184 0.042 3.20 0.10 648
(0.126) (0.107) (0.69)
Behavioral development −0.185 0.038 3.25 0.11 648
(0.113) (0.097) (0.68)
SUR test for joint significance (p-value)a 0.021 0.241
Panel A: Frequency of How Often Parents Did Reading Activities with their Child (standardized)
Mean Reading Activity −0.202 −0.156 — 0.06 648
(0.127) (0.113)
Let child hold the book −0.052 −0.160 2.82 0.03 648
(0.107) (0.117) (0.81)
Show child book parts −0.073 −0.142 3.05 0.04 648
(0.113) (0.119) (0.94)
Talk about the pictures in a book −0.218* −0.046 3.39 0.01 648
(0.126) (0.108) (0.75)
(0.119) (0.105) (0.81)
Underline words as you read −0.269** −0.299** 3.30 0.08 648
(0.106) (0.118) (0.83)
Show child that we read from left to right −0.112 −0.115 3.24 0.00 648
(0.127) (0.118) (1.01)
(0.129) (0.103) (0.99)
Panel B: Frequency of How Often Parents Did Literacy Activities with their Child (standardized)
Mean Literacy Activity −0.281** −0.144 — 0.06 648
(0.130) (0.122)
Helped child write his/her name −0.021 −0.041 2.56 −0.04 648
(0.126) (0.131) (0.94)
Pointed out letters on objects −0.195 −0.110 2.91 0.01 648
(0.126) (0.128) (0.92)
Sang a song or nursery rhyme with child −0.107 −0.099 2.87 −0.02 648
(0.126) (0.130) (0.92)
Practiced word sounds −0.309** −0.071 2.95 0.03 648
(0.119) (0.125) (0.92)
Said rhyming words with child −0.188 −0.101 2.77 0.04 648
(0.120) (0.123) (0.97)
Asked child about his or her day −0.357*** −0.203 3.45 0.03 648
(0.122) (0.123) (0.76)
Played a literacy learning game with child −0.148 −0.041 2.52 0.04 648
(0.118) (0.108) (1.01)
Helped child learn new words −0.245* −0.023 3.02 0.07 648
(0.126) (0.119) (0.90)
Worked on reading skills during family activities −0.293** −0.190 2.78 0.10 648
(0.118) (0.120) (0.95)
Helped child sound out a word −0.147 −0.074 2.91 0.05 648
(0.130) (0.122) (0.94)
Read to child −0.211* −0.188* 3.08 0.04 648
(0.119) (0.110) (0.87)
Panel C: Parents’ Understanding of How to Develop Skills for their Child (standardized)
Mean Skill Development −0.219* 0.034 — 0.11 648
(0.115) (0.101)
Literacy development −0.210* 0.009 3.28 0.06 648
(0.112) (0.115) (0.71)
Math development −0.184 0.042 3.20 0.10 648
(0.126) (0.107) (0.69)
Behavioral development −0.185 0.038 3.25 0.11 648
(0.113) (0.097) (0.68)
SUR test for joint significance (p-value)a 0.021 0.241

Notes: Each row represents a separate regression model (only the coefficients of the treatments status are reported). All regressions include controls for student characteristics (age, gender, race/ethnicity, low socioeconomic status, and Circle-1 test scores), parent characteristics (age and highest educational attainment), teacher characteristics (gender, race/ethnicity, years of experience, and hours of absence used), and preschool site fixed effects. Standard errors are clustered at the preschool site level. The omitted reference group in all regressions is the original texting program of three texts per week (i.e., FTG program).

aThe seemingly unrelated regression (SUR) test for joint significance is testing the joint significance of each program on all survey items in Table 5. Means and standard deviations are calculated from the underlying nonstandardized measures. FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

*p < 0.10; **p < 0.05; ***p < 0.01.

### Results on Child's Reading and Literacy Development

Although parents who received the Tip program were less likely to opt out of the texting intervention, they were also less likely to report to engage in parent–child reading and literacy activities and to understand their child's literacy development. These two effects on parent behavior could have an offsetting impact on their child's literacy development. Parents in the FTTTG program were more likely to opt out of the intervention and, as a result, should have, if anything, lower literacy scores compared with the original program.

Table 6 reports the treatment effects on children's literacy skill development. Although we find, on average, no treatment effects of either program compared with the FTG program (all coefficients are close to zero and statistically insignificant) on children's Circle-3 test, the effect of the texting programs is clearly dependent on the pre-intervention literacy skill distribution of the child. In particular, for children in the lowest quarter of the pretreatment Circle-1 literacy assessment, providing parents the Tip program led to lower literacy scores by 0.19 SD compared with the original program. Point estimates suggest that the literacy scores of children in the next higher two quarters increased by approximately 0.14 SD compared with the original texting program, but these effects are only marginally significant. Pooling these two middle quarters (not shown), we find a statistically significant effect of 0.13 SD (p < 0.001). The effects of the FTTTG program are also close to zero and statistically insignificant across all quarters of Circle-1 literacy scores.

Table 6.

Treatment Effects on Overall Circle-3 Language and Literacy Skills Assessment Test (standardized)

Full sample 0.019 −0.020 0.41 2920
(0.039) (0.041)
By quartiles:a
Quarter 1 (lowest) −0.186** −0.066 0.44 715
(0.089) (0.088)
Quarter 2 0.144* 0.060 0.40 728
(0.074) (0.087)
Quarter 3 0.136* 0.062 0.35 735
(0.077) (0.074)
Quarter 4 (highest) 0.003 −0.076 0.43 742
(0.064) (0.059)
Full sample 0.019 −0.020 0.41 2920
(0.039) (0.041)
By quartiles:a
Quarter 1 (lowest) −0.186** −0.066 0.44 715
(0.089) (0.088)
Quarter 2 0.144* 0.060 0.40 728
(0.074) (0.087)
Quarter 3 0.136* 0.062 0.35 735
(0.077) (0.074)
Quarter 4 (highest) 0.003 −0.076 0.43 742
(0.064) (0.059)

Notes: Each row represents a separate regression model (only the coefficients of the treatments status are reported). All regressions include controls for student characteristics (age, gender, race/ethnicity, low socioeconomic status, and Circle-1 test scores), parent characteristics (age and highest educational attainment), teacher characteristics (gender, race/ethnicity, years of experience, and hours of absence used), and preschool site fixed effects. Standard errors are clustered at the preschool site level. The omitted reference group in all regressions is the original texting program of three texts per week (i.e., FTG program). FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

aQuarters are based on student's Circle-1 tests prior to the intervention.

*p < 0.10; **p < 0.05.

The effects on the overall literacy scores reported in table 6 appear not to be driven by the score in any particular subtask. Table 7 shows treatment effects of the Tip and FTTTG programs for the various subtasks of Circle-3 language and literacy skills assessment test: (1) rapid letter naming, (2) rapid vocabulary, and (3) phonological awareness. Phonological awareness in turn can further be disaggregated into: (4) rhyming, (5) alliteration, and (6) syllabication. We report estimates across quarters of the Circle-1 literacy scores. Estimates for the full Circle-3 sample (not shown) are close to zero and statistically insignificant.24 We only find a significant negative effect for the Tip program for children in the lowest quarter on phonological awareness in comparison to the original program, which, in turn, appears to be driven by rhyming and syllabication. The remainder of subtasks in the lowest quarter have negative, but small and statistically insignificant point estimates. The literacy gains of children in the higher two middle quarters appear to be driven by rapid vocabulary naming, alliteration, and syllabication. However, the majority of treatment effects on scores in the different subtasks are not significantly different from each other within a given quarter. Therefore, we ultimately cannot distinguish whether a text message with one activity had differential effects on different literacy skills compared to three texts per week. The effects of the FTTTG program on the subtasks are mostly statistically insignificant with a few marginally significant exceptions.

Table 7.

Treatment Effects on Circle-3 Language and Literacy Skills Assessment Test by Specific Sub-tests (standardized)

(1) Rapid letter naming Quarter 1 −0.077 0.060 29.53 0.35 715
(lowest) (0.092) (0.092) (14.20)
Quarter 2 0.087 −0.044 30.00 0.37 728
(0.084) (0.075) (12.13)
Quarter 3 0.086 0.049 30.90 0.32 735
(0.085) (0.085) (12.18)
Quarter 4 0.016 −0.065 37.58 0.33 742
(highest) (0.074) (0.073) (11.21)
(2) Rapid vocabulary naming Quarter 1 −0.107 −0.018 18.72 0.47 715
(lowest) (0.086) (0.078) (8.26)
Quarter 2 0.142** 0.154* 18.89 0.44 728
Quarter 2 (0.065) (0.081) (8.14)
Quarter 3 0.068 −0.004 20.29 0.38 735
Quarter 3 (0.072) (0.069) (7.20)
Quarter 4 0.029 −0.019 23.67 0.41 742
(highest) (0.085) (0.072) (8.12)
(3) Phonological awarenessa Quarter 1 −0.270*** −0.204* 18.44 0.30 715
(lowest) (0.093) (0.108) (4.72)
Quarter 2 0.124 0.036 18.71 0.27 728
Quarter 2 (0.084) (0.100) (4.50)
Quarter 3 0.179** 0.106 19.40 0.23 735
Quarter 3 (0.090) (0.080) (3.85)
Quarter 4 −0.038 −0.103 21.78 0.25 742
(highest) (0.068) (0.067) (3.62)
(4) Rhyming Quarter 1 −0.322*** −0.222* 6.72 0.16 715
(lowest) (0.108) (0.127) (1.94)
Quarter 2 0.093 −0.034 6.86 0.15 728
Quarter 2 (0.105) (0.102) (1.87)
Quarter 3 0.041 0.044 7.02 0.15 735
Quarter 3 (0.108) (0.086) (1.80)
Quarter 4 −0.060 −0.151* 7.80 0.13 742
(highest) (0.083) (0.086) (1.43)
(5) Alliteration Quarter 1 −0.103 −0.028 5.07 0.23 715
(lowest) (0.095) (0.107) (1.70)
Quarter 2 0.196** 0.101 5.09 0.20 728
Quarter 2 (0.088) (0.104) (1.64)
Quarter 3 0.162 0.132 5.35 0.17 735
Quarter 3 (0.101) (0.085) (1.55)
Quarter 4 0.003 −0.045 6.10 0.20 742
(highest) (0.070) (0.073) (1.26)
(6) Syllabication Quarter 1 −0.200* −0.229* 5.69 0.23 715
(lowest) (0.105) (0.126) (1.91)
Quarter 2 −0.014 −0.033 5.69 0.18 728
Quarter 2 (0.090) (0.108) (1.68)
Quarter 3 0.185** 0.057 5.87 0.12 735
Quarter 3 (0.083) (0.089) (1.61)
Quarter 4 −0.018 −0.030 6.37 0.14 742
(highest) (0.073) (0.081) (1.27)
(1) Rapid letter naming Quarter 1 −0.077 0.060 29.53 0.35 715
(lowest) (0.092) (0.092) (14.20)
Quarter 2 0.087 −0.044 30.00 0.37 728
(0.084) (0.075) (12.13)
Quarter 3 0.086 0.049 30.90 0.32 735
(0.085) (0.085) (12.18)
Quarter 4 0.016 −0.065 37.58 0.33 742
(highest) (0.074) (0.073) (11.21)
(2) Rapid vocabulary naming Quarter 1 −0.107 −0.018 18.72 0.47 715
(lowest) (0.086) (0.078) (8.26)
Quarter 2 0.142** 0.154* 18.89 0.44 728
Quarter 2 (0.065) (0.081) (8.14)
Quarter 3 0.068 −0.004 20.29 0.38 735
Quarter 3 (0.072) (0.069) (7.20)
Quarter 4 0.029 −0.019 23.67 0.41 742
(highest) (0.085) (0.072) (8.12)
(3) Phonological awarenessa Quarter 1 −0.270*** −0.204* 18.44 0.30 715
(lowest) (0.093) (0.108) (4.72)
Quarter 2 0.124 0.036 18.71 0.27 728
Quarter 2 (0.084) (0.100) (4.50)
Quarter 3 0.179** 0.106 19.40 0.23 735
Quarter 3 (0.090) (0.080) (3.85)
Quarter 4 −0.038 −0.103 21.78 0.25 742
(highest) (0.068) (0.067) (3.62)
(4) Rhyming Quarter 1 −0.322*** −0.222* 6.72 0.16 715
(lowest) (0.108) (0.127) (1.94)
Quarter 2 0.093 −0.034 6.86 0.15 728
Quarter 2 (0.105) (0.102) (1.87)
Quarter 3 0.041 0.044 7.02 0.15 735
Quarter 3 (0.108) (0.086) (1.80)
Quarter 4 −0.060 −0.151* 7.80 0.13 742
(highest) (0.083) (0.086) (1.43)
(5) Alliteration Quarter 1 −0.103 −0.028 5.07 0.23 715
(lowest) (0.095) (0.107) (1.70)
Quarter 2 0.196** 0.101 5.09 0.20 728
Quarter 2 (0.088) (0.104) (1.64)
Quarter 3 0.162 0.132 5.35 0.17 735
Quarter 3 (0.101) (0.085) (1.55)
Quarter 4 0.003 −0.045 6.10 0.20 742
(highest) (0.070) (0.073) (1.26)
(6) Syllabication Quarter 1 −0.200* −0.229* 5.69 0.23 715
(lowest) (0.105) (0.126) (1.91)
Quarter 2 −0.014 −0.033 5.69 0.18 728
Quarter 2 (0.090) (0.108) (1.68)
Quarter 3 0.185** 0.057 5.87 0.12 735
Quarter 3 (0.083) (0.089) (1.61)
Quarter 4 −0.018 −0.030 6.37 0.14 742
(highest) (0.073) (0.081) (1.27)

Notes: All regressions include controls for student characteristics (age, gender, race/ethnicity, and low socioeconomic status), parent characteristics (age and highest educational attainment), teacher characteristics (gender, race/ethnicity, years of experience, and hours of absence used), and preschool site fixed effects. Standard errors are clustered at the preschool site level. The omitted reference group in all regressions is the original texting program of three texts per week (i.e., FTG program). Quarters are based on student's Circle-1 tests prior to the intervention. FTTTG = FACT-TIP-TIP-TIP-GROWTH; FTG = FACT-TIP-GROWTH.

aPhonological awareness is a composite score of the following assessments: rhyming, alliteration, and syllabication. Means and standard deviations (SD) are calculated from the underlying nonstandardized measures.

*p < 0.10; **p < 0.05; ***p < 0.01.

To put the effect sizes of the Tip program into context, the estimated effect on the phonological awareness score, for instance, translates to an average 1.13-point difference between students in the Tip and FTG programs. This corresponds to roughly 16.1 percent (7 points) of the expected learning gains between the English Circle-1 and Circle-3 for 4- to 4.5-year-old students, according to the Circle Technical Manual.

In addition to testing the effects of the programs on reading and literacy development, we investigated whether the Tip and FTTTG programs had differential impacts in comparison with the original program on math Circle-3 scores and attendance. Although the program did not explicitly target math development or attendance, we may expect effects on other student outcomes. Math development may be affected by changes in allocation of attention between education domains or by complementarities between math and literacy. Attendance may be affected if parents perceived the text messages as communications from (and therefore felt more closely connected to) the schools. In that case, the effects on literacy scores may be mediated by changes in attendance. We estimated our preferred specification overall and within quarters using overall Circle-3 math scores, attendance rates (days attended of days enrolled), and chronic absenteeism (more than 10 percent absent of all days enrolled).25 Overall, the results suggest that our programs did not differentially affect math scores and attendance. Most of our estimates are relatively small and insignificant at 10-percent significance level. The only exception is the estimate of the overall effect of the Tip program on the attendance rate (–0.03 percentage points, p < 0.10).

### Robustness Checks

In order to further probe our main results, we assess the robustness of our results to the systematic attrition differences across programs within quarters of the baseline literacy skill distribution. To understand the scope of the potential bias, we estimate the effects of the programs on the overall literacy score including children with missing scores. We estimate nine different regressions for the Circle-3 sample and within the second and third quarters—those that exhibited differential attrition rates, as shown in table 3 (panel A). In each regression, we impute missing scores with a different decile (1st to 9th decile) of the observed distribution in the respective sample. All models include site fixed effects and the full set of covariates.

#### Attrition Sensitivity Analysis for Circle-3 Literacy Test Score

Figure 2.
Attrition Sensitivity Analysis for Circle-3 Literacy Test Score

Notes: FTTTG = FACT-TIP-TIP-TIP-GROWTH.

Figure 2.
Attrition Sensitivity Analysis for Circle-3 Literacy Test Score

Notes: FTTTG = FACT-TIP-TIP-TIP-GROWTH.

Figure 2 shows these results. The dashed lines correspond to the baseline effects of the Tip (light gray) and FTTTG (dark gray) programs, respectively, not including children with missing scores. The solid lines correspond to the effects with missing scores imputed at a given decile of the observed distribution in the respective sample. As shown in the top panel, the results for the Tip and FTTTG programs in the Circle-3 sample are robust, but this is not surprising, given that we did not find a significant impact of the treatments on attrition for these samples. However, the estimated effects of the Tip program on the literacy score in the second and third quarter of the baseline skill distribution are not as robust. Recall that the Tip program appeared to have increased the likelihood of missing any Cirlce-3 scores by 4.4 percentage points in the second quarter and decreased the likelihood by 4.7 percentage points in the third quarter. In the second quarter, the estimated effect of the Tip program increases, assuming higher Circle-3 literacy scores for those children with missing values. When we assign all missing values to have scores at the 5th decile, the estimates are significantly different from zero. Conversely, in the third quarter, the estimates decrease, assuming higher Circle-3 literacy scores. The estimates are only significant at the first three deciles.

## 5.  Conclusion

This study analyzes content and frequency of a text messaging program aimed at supporting parent engagement in the literacy development of preschoolers. Overall, our findings provide evidence that text messaging programs can include too little and too much information. The original program of three texts per week was more effective at changing parent-reported behaviors and increasing learning for lower-achieving children than the Tip program with one text per week. This finding suggests that general information and encouragement are particularly useful for parents of lower-achieving children by highlighting the importance of the respective skill and by providing immediate gratification to parents. York, Loeb, and Doss (2019) estimated that the original program in comparison to a placebo program increased literacy development of children below the median of the base line skill distribution by 0.31 SD. We find that the original program is 0.19 SD more effective in supporting the literacy development of children in the lowest quarter of the base line skill distribution than a one tip per week program. Taken at face value, these two results suggest that one activity alone may still increase child literacy development by approximately 0.12 SD for lower-performing children compared to no treatment. No group that we identified benefited from five texts per week in the FTTTG program relative to the other two programs with fewer texts. Moreover, an increased number of texts led to greater program attrition and lower self-reported parent engagement. These results indicate that providing additional TIP messages with more parent–child literacy activities exacerbates cognitive demand or that more texts become bothersome to parents, thus leading some parents to opt out and to stop reaping the benefits of the text messages.

For parenting programs, as well as for other interventions aimed at changing adult behavior, it is easy to assume that more is better. Recent programs—some but not all using text-messaging to remind and provide information—have shown that light-touch interventions can have large effects, effects that are, in many cases, quite a bit larger than more intensive traditional programs. These light-touch programs commonly provide easy-to-implement suggestions over extended periods of time. Yet even in these light-touch programs, the balance between too much and too little can be quite salient. Our results point clearly to the possibility of too much information or too many text messages. This may even be true for information that recipients welcome and that help them reach goals that they prioritize. As such, schools should also consider how such light-touch behavioral programs may interact with additional text messages that schools are sending and with other forms of school communication. If schools already inform parents frequently about events, deadlines, absences, and so forth, via texts, they might inundate parents, and thus run the risk of losing parents’ attention even faster.

These results are in line with those of studies about text messaging interventions in other contexts. Cunha et al. (2017) investigate the effects of a text messaging program to parents of ninth-grade students in Sao Paolo, Brazil, that provides activities aimed to increase attendances and grade point average. Similar to ours, that program also used a fact-activity-growth model. The authors compare the effects of one (activity), two (fact and activity), and three texts per week (fact, activity, and growth). They find that whereas for attendance the effects plateau after two texts, three texts have the largest impact on grade point average. However, the authors cannot speak to a potential decrease of utility in response to higher numbers of text messages (e.g., five or more texts). Pop-Eleches et al. (2011), in contrast, find evidence of too many text messages in an intervention aimed to improve adherence to antiretroviral treatment. Whereas a weekly frequency of reminders (one text per week) had a significant positive on adherence, participants who received reminders daily (seven texts per week) showed no difference with the participants in the control group. In their setting, it is unclear if a more moderate increase in text messages (such as two or three texts per week) would have yielded larger effects than sending only one text per week. These results are in line with those of a meta-study of text messaging–based programs for health promotion (Head et al. 2013). The authors find that programs with low text frequencies have moderate effects, whereas programs with high frequencies have no significant effects. Maybe not surprisingly, they find that programs where recipients can choose the frequency or programs with decreasing frequencies have the largest effects. Note that these findings are not based on experimental variations of text message frequency but on cross-study comparisons. Taken together with these studies, our results show that there is a positive relationship between the number of texts and the effectiveness of text messaging programs for lower frequencies. Yet, additional texts appear to have a decreasing marginal effectiveness that can lead to program ineffectiveness.

Optimal program design includes a variety of design choices. Our study provides evidence that content and frequency of text messages matter. Other studies highlight the importance of additional program features, such as the day of the week the texts are sent (Cortes et al. 2019), the timing of text delivery (Cunha et al. 2017), the consistency of the delivery (Cunha et al. 2017; Gallego, Malamud, and Pop-Eleches 2017), and the appropriateness of the activity difficulty (Doss et al. 2019). Finally, in addition to the results presented here, Fricke, Kalogrides, and Loeb (2018) provide evidence that text messages with more complex language and programs with only activities, compared with programs that scaffold activities with context and encouragement, increase the likelihood that parents opt out. Moreover, our study, along with Cortes et al. (2019) and Doss et al. (2019), shows that design choices influence different parents differently. Hence, all parenting programs, including those not based in text messages, should carefully consider how advice and support is provided to parents with varying needs, as seemingly small design choices can lead to different results. Text messaging programs, with their low cost, ease of experimentation, and flexibility, are particularly well-suited for evidence-based design choices and optimal program differentiations.

## Acknowledgments

We thank Greg Duncan, Carolyn Heinrich, Tatiana Melquizo, Emily Penner, Lori Taylor, and seminar and conference participants at Stanford University, Texas A&M University, the University of California at Irvine, the University of Southern California, Vanderbilt University, the University of Fribourg, and the Association for Education Finance and Policy for helpful feedback. Erika Byun, Stephanie Gullo, and J. B. Horsley provided outstanding research assistance. Any errors are attributable to the authors. Institutional support from the Annenberg Institute at Brown University, Stanford University, and Texas A&M University are also gratefully acknowledged.

## Notes

1.

For example, Hart and Risley (1995) estimate that children from low-income families hear about 30 million fewer words at the age of 4 than children from high-income families.

2.

Such interventions have been demonstrated to positively affect school and class attendance of students (Groot et al. 2017; Robinson et al. 2018; Rogers and Feller 2018; Bergman and Chan 2021), the number of course credits earned in high school (Kraft and Rogers 2015), Free Application for Federal Student Aid completion (Page, Castleman, and Meyer 2020), chronic absenteeism and parental engagement (Smythe-Leistico and Page 2018), and college enrollment rates (Castleman and Page 2015, 2016). Moreover, these interventions have been particularly effective for children and parents from low-income backgrounds (Bergman 2015; Castleman and Page 2015, 2016; Bergman and Chan 2021). See also Gabraix (2017) for an overview and a theoretical approach to behavioral inattention in behavioral economics.

3.

See York, Loeb, and Doss (2019) for a detailed discussion of the different informational and behavioral barriers addressed with this texting program.

4.

To recruit parents for the study, we built on the district's existing school registration process for prekindergarten enrollment. When parents registered their children for preschool—a process that all parents must go through—they were invited to receive text messages and participate in the study. A study participation form, which included active consent and was vetted by the district, was available in both English and Spanish and was included in the district's preschool registration packet. Parents could choose to opt out of their program at any time during the school year.

5.

The text messages were not necessarily aligned with the day-to-day classroom activities nor were teachers actively informed about the content of the text messages and treatment status of parents.

6.

The text messages draw on research on literacy development (e.g., Lonigan and Shanahan 2009), parenting practices (e.g., Reese, Sparks, and Leyva 2010), and behavior change strategies (e.g., Abraham and Michie 2008). All text messages are couched in positive parenting practices with the goal of making the activities fun and engaging for both parent and child. We consulted Parent Management Training (Patterson, Reid, and Dishion 1992), Incredible Years (Webster-Stratton 1992), Triple P-Positive Parenting Program (Sanders et al. 2000), VIPP-SD intervention (Van Zeijl et al. 2006), and Family Check-Up (Gardner, Burton, and Klimes 2006). See York, Loeb, and Doss (2019) for a description of the original text development process.

7.

The Dallas ISD encompasses the cities of Dallas, Cockrell Hill, Seagoville, Addison, and Wilmer, and parts of Carrollton, Cedar Hill, DeSoto, Duncanville, Farmers Branch, Garland, Grand Prairie, Highland Park, Hutchins, Lancaster, and Mesquite.

8.

Results are available upon request.

9.

Studies have suggested that literacy-only curriculum may both increase children's literacy skills and enable them to more readily acquire math skills (e.g., Purpura et al. 2011). Thus, we also asked parents about supporting their child's math skills.

10.

Students are classified as economically disadvantaged if they qualify for free or reduced-price lunches or if other district-specific criteria apply.

11.

In the Circle-3 test sample, 31.23 and 66.54 percent of students were assessed in the English- and Spanish-language version of Circle, respectively. A small percentage of students (i.e., 2.23 percent; 65 students) in the Circle-3 test sample was assessed in both languages. For students who had both English and Spanish test scores, the higher score was used.

12.

In Appendix A (available in a separate online appendix that can be accessed on Education Finance and Policy’s Web site at https://doi.org/edfp_a_00304) we provide further details on each assessment, the administration of the assessments, and examples of each assessment.

13.

Only students who took the English-version of Circle were given the onset-rime task of the phonological awareness test. For this reason, we only analyzed three of the four subtasks. We standardized the phonological awareness composite score within language to take into account the fact that the English composite score includes onset-rime task and the Spanish composite score does not.

14.

We launched in the Dallas ISD on 16 November 2015 and the intervention ended on 24 June 2016.

15.

We excluded 359 families with no (218) or incomplete (141) Circle-1 scores, and 194 families with complete Circle-1 scores but no (138) or incomplete (56) Circle-3 scores. Of those that had not been assessed, we estimate 127 families had left the district in fall prior to the Circle-1 assessment date, and an additional 112 families had left the district in spring prior to the Circle-3 date. We estimate these numbers as the number of families with no test data who were not enrolled the entire spring or fall because we only know the number of days families were enrolled in prekindergarten in each term. In turn, 91 and 26 families were not assessed with Circle-1 and Circle-3 for other reasons, respectively.

16.

The benchmarks for the Spanish version of the test differs from those of the English version because they correspond to the 16th percentile of the version-specific score distribution. Moreover, the Spanish alphabet has twenty-seven letters compared with twenty-six letters in the English alphabet.

17.

We did not find any effects on Circle-3 math scores and student absences. Results are available upon request.

18.

Different model specifications do not change our results. Table B.3 in the online appendix shows the results with different specifications: estimates excluding the Circle-1 literacy scores from the set of covariates (shown in model 2), estimates only conditional on pre-school site fixed effects (shown in model 3), and estimates including a full set of covariates but excluding site fixed effects and without clustered standard errors (shown in model 4). Across all model specifications, the results are very similar.

19.

Similar balancing checks in each quarter of the Circle-1 literacy score distribution also suggest that randomization was successful within quarters. Results are available upon request.

20.

Children are included in the Circle-3 test sample if they have test scores in both Circle-1 and Circle-3 assessments.

21.

Disaggregating the attrition indicator suggests attrition is driven by families leaving the district in spring. The results are available upon request.

22.

Parents could answer the questions: “When you received Ready4K texts, did you READ them?” and “Did you USE the information in Ready4K texts?” these two questions were on a four-point Likert scale (never, sometimes, most of the time, always), the question “How HELPFUL was the information in Ready4K texts?” was on a four-point Likert scale (not helpful, a little helpful, helpful, very helpful), the question “Was the NUMBER of Ready4K texts that you received each week not enough, too many or just right?” was on a three-point Likert scale (not enough, too many, just right), and the question “To what extent would you DISCOURAGE or RECOMMEND Ready4K texts to other parents?” was on a three-point Likert scale (discourage, neither discourage nor recommend, recommend). Answers are standardized to have mean zero and standard deviation one.

23.

In panel A, parents could answer the question: “When you READ to your child, HOW OFTEN do you do the following things?” this question was on a four-point Likert scale (never, sometimes, often, always). In panel B, parents were asked to answer: “Last week, HOW MANY TIMES did you do each of the following READING RELATED activities with your child?” This question was on a four-point Likert scale (not at all, once or twice, 3 or 4 times, more than 4 times). In panel C, parents could answer the question: “How much do you AGREE with each of the following statements? I know what I can do to help my child develop/improve …?” this question was on a four-point Likert scale (strongly disagree, disagree, agree, strongly agree). All answers were standardized to have mean zero and standard deviation one.

24.

These results are available on request.

25.

The results are available upon request.

## REFERENCES

Abraham
,
Charles
, and
Susan
Michie
.
2008
.
A taxonomy of behavior change techniques used in interventions
.
Health Psychology
27
(
3
):
379
387
.
Aos
,
Steve
,
Roxanne
Lieb
,
Jim
Mayfield
,
Marna
Miller
, and
Annie
Pennucci
.
2004
.
Benefits and costs of prevention and early intervention programs for youth
.
Olympia, WA
:
Washington State Institute for Public Policy
.
Bergman
,
Peter.
2015
.
Parent–child information frictions and human capital investment: evidence from a field experiment
.
Munich, Germany
:
CESifo Working Paper
No.
5391
.
Bergman
,
Peter
, and
Eric W.
Chan
.
2021
.
Leveraging parents through low-cost technology: The impact of high-frequency information on student achievement
.
Journal of Human Resources
56
(
1
):
125
-
-158
.
,
Robert H.
,
Robert F.
Corwyn
,
Margart
Burchinal
,
Harriette P.
, and
Cynthia G.
Coll
.
2001
.
The home environments of children in the United States part II: Relations with behavioral development through age thirteen
.
Child Development
72
(
6
):
1868
1886
.
Castleman
,
Benjamin L.
, and
Lindsay C.
Page
.
2015
.
Summer nudging: Can personalized text messages and peer mentor outreach increase college going among low-income high school graduates?
Journal of Economic Behavior & Organization
115
:
144
160
.
Castleman
,
Benjamin L.
, and
Lindsay C.
Page
.
2016
.
Freshman year financial aid nudges: An experiment to increase FAFSA renewal and college persistence
.
Journal of Human Resources
51
(
2
):
389
415
.
Cortes
,
Kalena E.
,
Hans D. U.
Fricke
,
Susanna
Loeb
,
David S.
Song
, and
Benjamin N.
York
.
2019
.
When behavioral barriers are too high or low—How timing matters for parenting interventions
.
NBER Working Paper No. 25964
.
Cunha
,
Nina
,
Guilherme
Lichand
,
Ricardo
, and
Eric
Bettinger
.
2017
.
What is it about communicating with parents?
Unpublished paper
,
Stanford University
.
Doss
,
Christopher
,
Erin M.
Fahle
,
Susanna
Loeb
, and
Benjamin N.
York
.
2019
.
Supporting parenting through differentiated and personalized text messaging: Testing effects on learning during kindergarten
.
Journal of Human Resources
54
(
3
):
567
603
.
Duncan
,
Greg J.
,
Jens
Ludwig
, and
Katherine A.
Magnuson
.
2010
.
Child development
. In
Targeting investments in children: Fighting poverty when resources are limited
,
edited by
Phillip B.
Levine
and
David J.
Zimmerman
, pp.
27
58
.
Chicago
:
University of Chicago Press
.
Fricke
,
Hans
,
Demetra
Kalogrides
, and
Susanna
Loeb
.
2018
.
It's too annoying: Who drops out of educational text messaging programs and why
.
Economics Letters
173
:
39
43
.
Gabraix
,
Xavier
2017
.
Behavioral inattention
.
NBER Working Paper No. 24096
.
Gallego
,
Francisco
,
Ofer
Malamud
, Cristian Pop-Eleches.
2017
.
Parental monitoring and children's internet use: The role of information, control, and cues
.
NBER Working Paper No. 23982
.
Gardner
,
Frances
,
Jennifer
Burton
, and
Ivana
Klimes
.
2006
.
Randomized controlled trial of parenting intervention in the voluntary sector for reducing child conduct problems: Outcomes and mechanisms of change
.
Journal of Child Psychology and Psychiatry
47
(
11
):
1123
1132
.
Groot
,
Bibi
,
Michael
Sanders
,
Todd
Rogers
, and
Emily
Bloomenthal
.
2017
.
I get by with a little help from my friends: Two field experiments on social support and attendance in further education colleges in the UK
.
Available
Accessed 25 August 2020
.
Hart
,
Betty
, and
Todd R.
Risley
.
1995
.
Meaningful differences in the everyday experience of young American children
, 1st edition.
Baltimore, MD
:
Paul H. Brookes Publishing
.
,
Katharine J.
,
Seth M.
Noar
,
Nicholas T.
Iannarino
, and
Nancy G.
Harrington
.
2013
.
Efficacy of text messaging-based interventions for health promotion: A meta-analysis
.
Social Science and Medicine
97
:
41
48
.
Heckman
,
James J.
2006
.
Skill formation and the economics of investing in disadvantaged children
.
Science
312
(
5782
):
1900
1902
.
Karoly
,
Lynn A.
,
Peter W.
Greenwood
,
Susan S.
Everingham
,
Jill
Houbé
,
M.
Rebecca Kilburn
,
C.
Peter Rydell
,
Matthew
Sanders
, and
James
Chiesa
.
1998
.
Investing in our children: What we know and don't know about the costs and benefits of early childhood interventions
.
Santa Monica, CA
:
RAND
.
Kraft
,
Matthew A.
, and
Todd
Rogers
.
2015
.
The underutilized potential of teacher-to-parent communication: Evidence from a field experiment
.
Economics of Education Review
47
:
49
63
.
Lonigan
,
Christopher J.
, and
Timothy
Shanahan
.
2009
.
Developing early literacy: Report of the National Early Literacy Panel
.
Washington, DC
:
National Institute for Literacy
.
Page
,
Lindsay C.
,
Benjamin
Castleman
, and
Katharine
Meyer
.
2020
.
Customized nudging to improve FAFSA completion and income verification
.
Educational Evaluation and Policy Analysis
42
(
1
):
2
21
.
Patterson
,
Gerald R.
,
John B.
Reid
, and
Thomas J.
Dishion
.
1992
.
Antisocial boys: A social interactional approach
.
Eugene, OR
:
Castalia Publishing Co
.
Pop-Eleches
,
Cristian
,
Harsha
Thirumurthy
,
James P.
Habyarimana
,
Joshua G.
Zivin
,
Markus P.
Goldstein
,
Damien de
Walque
,
Leslie
MacKeen
,
Jessica
Harberer
,
Sylvester
Kimaiyo
,
John
Sidle
,
Duncan
Ngare
, and
David R.
Bangsberg
.
2011
.
Mobile phone technologies improve adherence to antiretroviral treatment in a resource-limited setting: A randomized controlled trial of text message reminders
.
AIDS
25
(
6
):
825
834
.
Purpura
,
David J.
,
Laura E.
Hume
,
Darcey M.
Sims
, and
Christopher J.
Lonigan
.
2011
.
Early literacy and early numeracy: The value of including early literacy skills in the prediction of numeracy development
.
Journal of Experimental Child Psychology
110
(
4
):
647
658
.
Reese
,
Elaine
,
Alison
Sparks
, and
Diana
Leyva
.
2010
.
A review of parent interventions for preschool children's language and emergent literacy
.
Journal of Early Childhood Literacy
10
(
1
):
97
117
.
Robinson
,
Carly D.
,
Monica G.
Lee
,
Eric
Dearing
, and
Todd
Rogers
.
2018
.
Reducing student absenteeism in the early grades by targeting parental beliefs
.
American Educational Research Journal
55
(
6
):
1163
1192
.
Rogers
,
Todd
, and
Avi
Feller
.
2018
.
Reducing student absences at scale by targeting parents’ misbeliefs
.
Nature Human Behavior
2
:
335
342
.
Sanders
,
Matthew R.
,
Carol
,
Lucy A.
Tully
, and
William
Bor
.
2000
.
The triple P-positive parenting program: A comparison of enhanced, standard, and self-directed behavioral family intervention for parents of children with early onset conduct problems
.
Journal of Consulting and Clinical Psychology
68
(
4
):
624
640
.
Smythe-Leistico
,
Ken
, and
Lindsay C.
Page
.
2018
.
Connect-text: Leveraging text–message communication to mitigate chronic absenteeism and improve parental engagement in the earliest years of schooling
.
Journal of Education for Students Placed at Risk
23
(
1–2
):
139
152
.
Stevenson
,
Harold W.
,
Chuansheng
Chen
, and
David H.
Uttal
.
1990
.
Beliefs and achievement: A study of black, white, and Hispanic children
.
Child Development
61
(
2
):
508
523
.
Texas Education Agency
.
2015
.
Texas prekindergarten guidelines
.
Available
Accessed 25 August 2020
.
Van Zeijl
,
Jantien
,
Judi
Mesman
,
Marinus H. Van
IJzendoorn
,
Marian J.
Bakermans–Kranenburg
,
Femmie
Juffer
,
Mirjam N.
Stolk
,
Hans M.
Koot
, and
Lenneke R.A.
.
2006
.
Attachment-based intervention for enhancing sensitive discipline in mothers of 1- to 3-year-old children at risk for externalizing behavior problems: A randomized controlled trial
.
Journal of Consulting and Clinical Psychology
74
(
6
):
994
1005
.
Webster-Stratton
,
Carolyn
1992
.
The Incredible Years: Basic program manual
.
Seattle, WA
:
The Incredible Years
.
York
,
Benjamin N.
,
Susanna
Loeb
, and
Christopher
Doss
.
2019
.
One step at a time: The effects of an early literacy text messaging program for parents of preschoolers
.
Journal of Human Resources
54
(
3
):
537
566
.