Journal Critique: Labor Market Returns to Early Childhood Stimulation in Jamaica

Matched samples, I've learnt, alongside random assignment are the key to a robust comparative research design. Gertler and colleagues (2014) longitudinal study was informative of covariates that can radically impact on learning outcomes. Below is an assignment from my post-grad studies, the headings are from the rubric as we were not assigned any specific risk of bias assessment tool.

journal critique early childhood longitudinal covariate matched sample

Research Questions

• Not clearly stated; cross-referenced with an earlier publication about the study, Gertler et al. (2013)

• Can high quality maternal child modeling improve parental investment and subsequent labour returns for the community?

• Does early childhood-development have greater impact in developing countries (compared to developed nations) because of lower stimulation at home prior to an intervention?

• What are the impacts of toddler psychosocial stimulation on young adult earnings?

• Will stimulation enable a stunted treatment group to catch up with a non stunted comparison group and compensate for psychological delays?

Population of Interest

• Non stunted comparison group

Population of the Study

• Growth stunted children living in poverty with their mothers in Kingston, Jamaica 1986-1987

Population of Inference

• Apx. 200 million children under the age of 5 living in developing countries at risk of not reaching their full potential as citizens

Intervention Conditions

Toddlers (1986


Children aged between 9 and 24 months

Intervention groups were children stunted in their growth (> 2 SD

Stratified by age and gender

Random assignment to one of 4 conditions:

psychosocial stimulation (n = 32)

psychosocial and nutrition (n = 32)

nutrition (n = 32)

control (n = 33)

NB. Nutritional supplement had no significant effect

 pooled groups

Treatment group = psychosocial + psychosocial and nutritional groups (n = 64)

Control group = nutrition + control groups (n = 65)

Tested for robustness

Plus 1 non stunted comparison group (n = 84)

•‘Sensible’ sample number in that at least 30 in each group (Tabachnick &

Fidel, 2013)