Determinants of Oral Reading Fluency among Learners in Kenya's Early Grades

1. INTRODUCTION

Kenya's primary education system has expanded access dramatically since the introduction of Free Primary Education in 2003, yet participation has not translated into commensurate gains in foundational literacy (Piper, Zuilkowski, & Mugenda, 2014). Citizen-led household assessments have repeatedly shown that a substantial share of children who complete several years of primary schooling cannot read a simple Grade 2–level story in English or Kiswahili (Uwezo, 2016; Usawa Agenda, 2023). Oral reading fluency (ORF) is correct words read per minute (CWPM), combined with accuracy, and has become the dominant metric for tracking this problem because it correlates strongly with reading comprehension and predicts later academic achievement (Piper & Zuilkowski, 2015; Platas, Perry, Piper, Sitabkhan, & Ketterlin-Geller, 2022).

Over the past decade, Kenya has been a proving ground for early-grade reading and numeracy interventions, most delivered as time-bound donor- or NGO-funded projects: the Primary Math and Reading Initiative (PRIMR), Tusome, the Kenya GPE Primary Education Development Project (PRIEDE), the Girls' Education Challenge (GEC) portfolio, the Health and Literacy Intervention (HALI), Literacy Boost, Bridge International Academies, the Developing Readers Study, Read-Aloud Now (RANA), and Wasichana Wetu Wafaulu (WWW), among others. Each closed with an end-of-project (endline) evaluation that re-administered an Early Grade Reading Assessment (EGRA) and, in several cases, its numeracy counterpart, EGMA, to representative samples of Grade 1–3 learners. Collectively, these fifteen endline reports constitute one of the richer, if fragmented, evidence bases in Sub-Saharan Africa on how ORF varies with who the child is (gender, age, grade-repetition history, pre-school attendance, home and parental background) and what the teacher does in the classroom.

What has been missing is a systematic bringing-together of this evidence. Individual project reports rarely reference one another's disaggregated findings, and academic syntheses tend to treat 'Kenya' as a single data point rather than as a decade-long, multi-project natural experiment. This article addresses that gap by asking: across Kenya's end-of-project EGRA evaluations conducted between 2014 and 2024, how consistently, and through which pathways, are child characteristics and teacher practice associated with oral reading fluency performance among Grade 1 and Grade 3 learners in English and, where reported, Mathematics?

Drawing on the reported statistical findings of fifteen independent end-of-project evaluations and twelve peer-reviewed studies, this article advances the following claim: oral reading fluency among Kenyan Grade 1–3 learners is systematically patterned by a cluster of child-level characteristics gender, pre-school attendance, and grade-repetition history most consistently reported but teacher instructional practice is the single characteristic most strongly and consistently associated with fluency gains across projects, at times outweighing fixed child or household attributes. Where projects deliberately altered teacher practice (coaching, structured lesson guides, print exposure, read-aloud protocols), fluency gains were larger and more consistently statistically significant than where projects manipulated only materials or access without changing instruction.

2. LITERATURE REVIEW

Measurement precedes explanation. Piper and Zuilkowski (2015) compared oral and silent EGRA administration in Kenya and found no significant difference in scores between modes but recommended oral assessment because it alone yields correct-words-per-minute data and permits diagnostic error analysis, the methodological basis on which every report reviewed in this article relies. Piper, Schroeder, and Trudell (2016) extended this by assessing 2,000 Kenyan children in up to three languages, showing that English decoding was often easier than mother-tongue decoding, yet English comprehension lagged behind Kiswahili and mother-tongue comprehension a language-of-instruction effect that complicates any simple reading of 'gender' or 'background' effects unless language is held constant, a caution this synthesis observes by reporting English and Kiswahili results separately where projects distinguish them.

On intervention effects, Piper, Zuilkowski, and Mugenda (2014) established in a randomised trial that PRIMR's combination of teacher coaching, instructional support, and pupil-book ratios produced statistically significant, if modest, first-year ORF gains, with larger effects in non-formal schools an early signal, corroborated repeatedly in the findings below, that teacher-facing inputs move fluency more reliably than pupil-facing inputs alone. Wawire (2020) traced this further upstream, arguing that Kenyan pre-service teacher-preparation curricula under-equip new teachers with the pedagogical content knowledge specific to early reading instruction, which helps explain why in-service coaching (rather than credentialing) recurs as the active ingredient across the evaluations synthesised here. At the county level, Cherobon and Chepsiror (2022) found that Tusome's reach in Nandi County correlated with improved textbook access and higher Grade 3 reading ability scores, consistent with the national Tusome endline patterns reported below.

Child-characteristic literature outside Kenya offers comparators. Kim (2022), studying Ethiopia's pre-primary expansion, found that preschool participation's association with literacy became positive and significant only after national-scale expansion, with little heterogeneity in gains by gender, urbanity, or parental literacy, suggesting that pre-school effects on later ORF may depend on programme scale and quality more than on fixed child attributes. Ahmed, Qin, and Lietz (2025), examining six Southeast Asian systems, similarly found parental support and preschool experience jointly predictive of reading literacy, reinforcing that home and school inputs interact rather than operate independently. Platas et al. (2022), using Kenyan longitudinal data, showed that school-entry literacy and mathematics skills predicted end-of-Grade-2 achievement even after controlling for socio-economic status, rural/urban residence, and parental literacy, evidence that child-level starting skills carry an independent signal beyond household background alone.

On gender specifically, Fonseca et al. (2023), pooling nineteen USAID-funded EGRA studies across fourteen countries including Kenya, found a consistent early female advantage in oral reading fluency that narrows, but does not reverse, as boys' scores catch up through the fluency distribution a pattern this synthesis observes directly in the RANA and Developing Readers Study findings. Pitchford, Chigeda, and Hubber (2019), working in Malawi with reference to the wider East African evidence, found that interactive technology reduced gender gaps specifically in early mathematics, indicating that the direction and magnitude of gender effects are outcome- and modality-specific rather than uniform across literacy and numeracy.

Finally, on home background, Brew, Atiso, and Osei (2021) found parental education positively associated with children's academic achievement across Sub-Saharan Africa, while Madegwa, Piliyesi, and Katundano (2019), studying Kakamega County, Kenya, found parental socio-economic background significantly associated with academic performance. Both studies, however, concern general academic performance rather than early-grade ORF specifically, a gap that foreshadows one of this article's central findings: Kenyan early-grade endline evaluations measure home and parental background far less consistently than they measure gender or teacher practice.

2. METHODS

3.1 Design

This article uses a narrative evidence-synthesis design. It draws exclusively on the statistical findings already computed and reported within each source evaluation; it does not re-analyse raw microdata, because unit-record datasets for most of the fifteen evaluations are not in the public domain. Where an evaluation reports a significance test, effect size, or percentage-point comparison, that figure is reproduced and attributed; where an evaluation reports only descriptive percentages, this is stated as such rather than treated as inferential evidence. This design choice is a deliberate limitation, discussed further in Section 6, and distinguishes the article from a primary quantitative study using chi-square or multinomial logistic regression on original survey data.

3.2 Selection Criteria

Reports were eligible for the findings if they met all of the following: (a) implemented in Kenya; (b) an end-of-project or endline evaluation of an early-grade literacy and/or numeracy intervention, rather than a stand-alone baseline or midline report considered in isolation; (c) used an EGRA or EGRA-type oral reading (or, where relevant, EGMA numeracy) assessment among Grade 1–3 learners, or reported findings disaggregated for these grades; (d) publicly documented and dated between 2014 and 2024; and (e) reported at least one child-characteristic or teacher-practice disaggregation of outcomes. Two entries required a documented exception to this last criterion: no independently published, standalone endline evaluation report could be located for a 'World Vision CESSP' project or for Bridge International Academies as an externally verified end-of-project evaluation. For these two rows, the closest available substitutes are used: the World Vision Kenya/Save the Children Literacy Boost baseline-endline design in Kitui and Makueni counties, and Bridge International Academies' own commissioned 2013–14 impact-evaluation report and both are flagged with an asterisk (*) in Tables 4.1 and 4.3 and interpreted with corresponding caution.

Note: * indicates a substituted or partially independent source (see Section 3.2). Reports are grouped by type in the Findings section: citizen-led national assessments (#1–4), government/donor-scale programme evaluations (#5–10), and NGO/private-sector or targeted-project evaluations (#11–15).

3.3 Analytic Approach

For each report, this synthesis records: (i) which of the six target characteristics (gender, age/grade, pre-school attendance, grade repetition, home/parental background, teacher practice) the report disaggregates outcomes by; (ii) the numerical finding as reported, including any significance test, p-value, effect size in standard deviations, or CWPM difference the original evaluators computed; and (iii) the assessment domain (English, Kiswahili, or Mathematics/EGMA) to which the finding applies. Findings are grouped into three tables of five reports each (Section 4) to keep the source of every reported statistic traceable to a specific project and citation.

3. RESULTS

Across the fifteen reports, gender and teacher practice are the two characteristics most frequently and most precisely disaggregated; pre-school attendance and grade repetition appear in fewer than half of the reports as an explicit ORF disaggregation (more often as a general programme-monitoring indicator); and home/parental background is disaggregated with numerical precision in only one of the fifteen reports at endline (Wasichana Wetu Wafaulu's disability sub-assessment being a partial exception), despite being a stated design objective in at least two others (Literacy Boost Kenya).

Table 4.1 illustrates the central tension this article investigates: national, population-level indicators (Uwezo 2015, 2016, 2021; Usawa Agenda 2023) show near-flat aggregate ORF outcomes across a decade despite dozens of concurrent interventions, while a tightly implemented, teacher-facing randomised intervention (PRIMR) produced statistically significant, if modest, gains within a single school year. This is consistent with the interpretation that project-level teacher-practice effects are real but have not aggregated into system-level change, rather than with the interpretation that teacher-practice effects are illusory.

Table 4.2 shows a similar pattern at larger scale. Tusome's national rollout more than doubled the share of Grade 2 learners reading at grade level over the life of the programme, tracking closely with rising teacher lesson-completion rates (#6–7) and an association between reported teacher-practice fidelity and ORF outcomes, though the publicly available reports do not isolate the two through formal regression. By contrast, PRIEDE's mathematics component shows that scaling teacher training in volume (#8) did not, by itself, prevent a measured decline in a specific numeracy subtask, reinforcing that the quality and content of teacher practice, not training exposure alone, is the operative variable. HALI (#10) provides the most direct experimental evidence in this group: exogenously increased classroom print exposure, a teacher-practice variable, produced statistically significant gains in fluency and comprehension.

Table 4.3 contains the two most granular child-characteristic findings in the entire dataset. The Developing Readers Study (#13) and RANA (#14) both report gender-disaggregated ORF gains under a defined teacher-practice change (structured remedial lesson guides; read-aloud instruction, respectively), and both show that girls benefited more than boys from the same practice change, evidence that teacher practice and gender interact rather than operate as independent, additive predictors. The Bridge International Academies figures (#12), while informative, come from a programme-commissioned evaluation that bundles school type, teacher practice, and household selection effects together and should not be read as isolating any single characteristic. The Literacy Boost Kenya entry (#11) illustrates a recurring limitation identified in Section 3.2: a report explicitly designed to examine home-background effects on reading, for which no matching endline disaggregation by that characteristic could be in the public domain.

4. DISCUSSION

The findings support three conclusions that connect back to the literature reviewed. First, teacher instructional practice variously operationalised as coaching and instructional support (PRIMR), lesson-guide fidelity (Developing Readers Study, Tusome), print exposure (HALI), or read-aloud protocol (RANA) is the characteristic most consistently associated with statistically significant ORF gains across the fifteen reports, echoing Piper, Zuilkowski, and Mugenda (2014) and Wawire's (2020) argument that pedagogical content knowledge, not access or materials alone, is the binding constraint on Kenyan early reading outcomes.

Second, gender is the child characteristic most frequently and precisely reported, and the direction of the effect is broadly consistent with Fonseca et al.'s (2023) multi-country finding of an early female fluency advantage: girls outperformed boys in the Developing Readers Study and RANA and, by design, benefited more from GEC-funded interventions. However, this synthesis also shows that gender effects are not fixed RANA’s finding that the same read-aloud practice narrowed, rather than reversed, an initial boys' advantage in one subtask, and Kim's (2022) finding of limited gender heterogeneity in preschool effects, both caution against treating gender as a uniformly-sized effect independent of the specific instructional practice or outcome measured.

Third, and most consequential for future evaluation design, pre-school attendance, grade repetition, and home/parental background are the least consistently measured characteristics in Kenya's endline evaluation record, despite Platas et al. (2022) and Brew, Atiso, and Osei (2021) showing these variables carry independent predictive weight in Kenyan and regional data respectively. Where reported at all, these characteristics tend to appear in baseline design documents (Literacy Boost Kenya) or programme-monitoring statistics (PRIEDE teacher counts) rather than as endline ORF disaggregations with significance tests. This is a structural gap in the evidence base rather than a substantive null finding, and it limits how confidently this synthesis, or any future one relying on the same public reports, can adjudicate between child-characteristic and teacher-practice explanations using existing secondary data alone.

5. CONCLUSION

The evidence reviewed here supports the claim in its core structure but with an important qualification. Teacher instructional practice is, across all three groupings of reports (national assessments and PRIMR; government/donor-scale programmes; NGO and targeted projects), the characteristic most consistently and most strongly associated with statistically significant oral reading fluency gains among Kenyan Grade 1–3 learners more consistently than any single fixed child characteristic. Gender is the child characteristic most reliably measured and shows a real, if practice-dependent, pattern favouring girls in several projects. Pre-school attendance and grade repetition appear as programme-design considerations more often than as measured ORF predictors at endline, and home/parental background, despite strong justification in the wider literature (Platas et al., 2022; Brew, Atiso, & Osei, 2021) for its inclusion, is the least consistently reported characteristic in Kenya's endline evaluation record.

The qualification, then, is not that child characteristics do not matter, but that Kenya's existing end-of-project evaluation infrastructure is not yet designed to measure most of them with the same rigour it applies to teacher-practice indicators and, to a lesser extent, gender. For the claim to be tested more definitively, future endline evaluations and any subsequent doctoral or academic analysis drawing on primary survey data, including chi-square tests of association and multinomial logistic regression models predicting fluency-performance categories, should standardise the collection and disaggregated reporting of pre-school attendance, grade-repetition history, and parental education and occupation alongside the gender and teacher-practice indicators already routinely collected. Until then, the most defensible conclusion from the existing Kenyan evidence base is that improving and standardising teacher instructional practice remains the most consistently evidenced lever for raising oral reading fluency in Kenya's early grades, while the role of specific child and household characteristics, though real, is less uniformly documented and therefore less precisely quantifiable from the published record alone.

These findings should be read against clear limitations. The fifteen reports use different assessment languages, grade-level samples, benchmark definitions, and statistical methods (t-tests, ANCOVA, propensity-score-matched multinomial logistic regression, and simple percentage-point comparison all appear across the set), which precludes pooling them into a single meta-analytic effect size. Two of the fifteen sources (#11, #12) are acknowledged substitutions or program-commissioned evaluations rather than fully independent endline reports, and are interpreted accordingly. Finally, because this is a synthesis of already-published statistics rather than a reanalysis of microdata, it cannot itself run the chi-square or multinomial logistic regression tests needed to directly test interaction effects between, say, gender and teacher practice; it can only report where the original evaluators did so.

DECLARATION

Conflict of interest

The authors declare no conflict of interest of any form regarding this work, and is, thus, their (own) original work done independently.

REFERENCES

African Population and Health Research Center. (2024). Developing Readers Study: Endline report. Nairobi: APHRC.

Ahmed, S. K., Qin, C., & Lietz, P. (2025). Parental support and preschool: Do they matter for primary students' reading literacy? Evidence from six Southeast Asian countries. Educational Research and Evaluation. https://doi.org/10.1177/17454999251324766

Brew, L., Atiso, K., & Osei, M. (2021). The influence of parental education on children's academic achievement in sub-Saharan Africa. International Journal of Educational Research, 109, 101788. https://doi.org/10.1016/j.ijer.2021.101788

Bridge International Academies. (2016). The Bridge Effect: A comparison of early grade learning gains in English and maths. 2013–14 impact evaluation results. Nairobi: Bridge International Academies.

Cherobon, V., & Chepsiror, E. (2022). Influence of Tusome early grade reading intervention on the reading abilities of learners in public primary schools in Nandi County, Kenya. East African Journal of Education Studies, 5(2), 61–68. https://doi.org/10.37284/eajes.5.2.655

Education Development Trust. (2023). Wasichana Wetu Wafaulu (WWW): GEC-Transition endline evaluation report. London: FCDO Girls' Education Challenge.

Fonseca, J., Bahrawar, L., Dubeck, M. M., Sitabkhan, Y., Cummiskey, C., & Unadkat, D. (2023). Girls have academic advantages and so do boys: A multicountry analysis of gender differences in early grade reading and mathematics outcomes. RTI Press. https://doi.org/10.3768/rtipress.2023.rr.0049.2305

Jukes, M. C. H., Turner, E. L., Dubeck, M. M., Halliday, K. E., Inyega, H. N., Wolf, S., Zuilkowski, S. S., & Brooker, S. J. (2016). Improving literacy instruction in Kenya through teacher professional development and text-message support: A cluster randomized trial. Journal of Research on Educational Effectiveness (Health and Literacy Intervention project report).

Kim, J. H. (2022). Preschool participation and students' learning outcomes in primary school: Evidence from national reform of pre-primary education in Ethiopia. International Journal of Educational Development, 94, 102659. https://doi.org/10.1016/j.ijedudev.2022.102659

Madegwa, B. M., Piliyesi, S. D. E., & Katundano, S. D. T. (2019). Parental socio-economic background and students' academic performance in public secondary schools in Ikolomani Sub-County, Kakamega County, Kenya. Journal of Education and Practice, 3(1), 15–35. https://doi.org/10.47941/jep.309

Moussa, W., Koester, B., & Alonge, O. (2018). Read-Aloud Now (RANA) impact evaluation report. Washington, DC: Education Policy and Data Center (EPDC) / USAID.

NORC at the University of Chicago. (2020). Tusome external evaluation: Endline report. Chicago: NORC / USAID.

Piper, B., & Zuilkowski, S. S. (2015). Assessing reading fluency in Kenya: Oral or silent assessment? International Review of Education, 61(2), 153–171. https://doi.org/10.1007/s11159-015-9470-4

Piper, B., Schroeder, L., & Trudell, B. (2016). Oral reading fluency and comprehension in Kenya: Reading acquisition in a multilingual environment. Journal of Research in Reading, 39(2), 133–152. https://doi.org/10.1111/1467-9817.12052

Piper, B., Zuilkowski, S. S., & Mugenda, A. (2014). Improving reading outcomes in Kenya: First-year effects of the PRIMR Initiative. International Journal of Educational Development, 37, 11–21. https://doi.org/10.1016/j.ijedudev.2014.02.006

Pitchford, N. J., Chigeda, A., & Hubber, P. J. (2019). Interactive apps prevent gender discrepancies in early-grade mathematics in a low-income country in sub-Saharan Africa. Developmental Science, 22(5), e12864. https://doi.org/10.1111/desc.12864

Platas, L. M., Perry, L., Piper, B., Sitabkhan, Y., & Ketterlin-Geller, L. (2022). School-entry predictors of lower primary reading and mathematics achievement in Kenya. Journal of Research in Reading, 45(3), 313–333. https://doi.org/10.1177/17454999221084414

PwC. (2017). Girls' Education Challenge Step Change Window: Endline evaluation report. London: UK Department for International Development.

RTI International. (2014). Improving reading outcomes in Kenya: First-year effects of the PRIMR Initiative (project endline report). Research Triangle Park, NC: RTI International / USAID–DFID.

RTI International. (2017). Tusome external evaluation: Midline report. Research Triangle Park, NC: RTI International / USAID–DFID.

Usawa Agenda. (2023). Foundational Literacy and Numeracy Assessment (FLANA) report 2023. Nairobi: Usawa Agenda.

Uwezo Kenya. (2015). Are our children learning? The state of education in Kenya in 2015 and beyond. Nairobi: Twaweza East Africa.

Uwezo Kenya. (2016). Are our children learning? Uwezo Kenya sixth learning assessment report. Nairobi: Uwezo.

Uwezo Kenya. (2021). Are our children learning? National learning assessment report. Nairobi: Uwezo / Twaweza East Africa.

Wawire, B. (2020). Promoting effective early grade reading: The case study of primary teachers' preparation programmes in Kenya. The Curriculum Journal, 32(2), 247–268. https://doi.org/10.1002/curj.69

World Bank. (2019). Kenya GPE Primary Education Development Project (PRIEDE): Implementation status and results report (P146797). Washington, DC: World Bank.

World Vision Kenya & Save the Children. (2013–2014). Literacy Boost Kenya: Baseline report and endline monitoring, Kalawa and Mutomo Area Development Programmes, Makueni and Kitui Counties. Nairobi: World Vision Kenya / Save the Children.