PRIDI 2013 Results

Background scales
Scale Creation

The PRIDI socio-emotional background scale was composed of multiple-choice questions regarding the children’s social and emotional development. The questions were answered by a parent or guardian.

The questions had four response categories: almost never (1); sometimes (2); often (3); and almost always (4):

  • Preliminary statistical analysis showed that respondents did not distinguish between the categories “almost never” and “sometimes.”
  • For the purpose of analyzing this scale, categories 1 and 2 were accordingly collapsed into a single category.

In addition, preliminary statistical analysis showed that question number 7 (PNG26G), asking whether the child cries when the parent or guardian leaves, did not correlate with the rest of the items, and consequently did not correlate with the total score on the scale. This item was therefore removed for the purpose of scaling and computing scores for this scale.

As was done with the development items, the data for the socio-emotional scale was scaled using partial credit modeling, with the discrimination parameters for all items fixed to 1.

As with the Engle scale (see below), the socio-emotional scale scores were standardized and placed on a more useful metric with a mean of 50 and standard deviation of 5.


List of Background Scales

Socio-Emotional Scale

The scales was composed based on questions related to children’s:

  • Relationships with other children and adults,
  • Willingness to participate in various activities with others,
  • Degree of independence in certain daily routines,
  • Response to specific situations.
Other scales
Child development scales - Scale creation

To achieve its goal of coverage of two age groups, the PRIDI instruments included a range of items that covered a series of tasks that the child being tested was asked to perform.

  • Depending on age, the child was administered one of two forms, with the two forms having overlapping items.
  • PRIDI relied on item response theory (IRT) scaling to combine the children’s responses and provide accurate estimates of proficiency on each of these scales.

The Engle scale focused on three major domains of skills in children’s development: cognitive, motor, and language skills.

  • Each of the two forms utilized (A and B) contained a set of common items, as well as a set targeted to the corresponding age group.
  • When scaling the items, the common items were used to create a single scale for both forms, hence their denomination as anchor items.
  • The answers to the items were scored “correct” or “incorrect,” in the case of dichotomous items, and “correct,” “partially correct,” or “incorrect” in the case of polytomously scored items.

Descriptive statistics were computed for each item.

  • The means for dichotomous items represented the proportion of children that answered the items correctly.
  • The means for the polytomous items represented the average score on the item.
  • A mean closer to the maximum score indicated an easy item.
  • A mean closer to zero indicated a difficult item.

Items were calibrated onto a single scale using a one-parameter IRT model where the probability of a response is modeled as a function of the difficulty of the item and the ability of the person.

  • The discrimination parameter was fixed at 1, and scores were calculated using a weighted maximum likelihood estimation procedure.
  • Calibration and initial assignment of scores was conducted using Parscale 4.0.
  • For the purpose of estimating the item difficulties, sampling weights were used in such a way that each country contributed equally to the difficulty of the items, and within each country, students from each of the two age groups (Form A and Form B) also contributed equally. 

Finally, the scores were standardized and placed on a more useful metric with a mean of 50 and standard deviation of 5.

  • The mean and the standard deviation were obtained by combining all countries (4), and age groups (2), and using sampling weights that take into account the sampling within the country, but also equalize the contribution of each of these 8 groups.
  • This standardization was done separately for each scale.


List of Development scales




Mean of  the cognitive, motor, and vocabulary scales

Overview of key study results

Child development in Latin America has been unequal

  • Inequality in results appeared as early as 24 months, the age of PRIDI’s youngest participants, and increased with age.
  • There was variation in the inequality. For example, correlations of the results of the Engle scale with the socio-economic characteristics of the home and maternal education were stronger for cognition and language and communication than for motor development.


The nurturing environment played a key role

The environment within which children developed – the nurturing environment – was important in all domains of child development measured by the Engle scale, although stronger associations appeared for cognition, language and communication, and socio-emotional development.

  • Gaps between the development of children in the top and low extremes in these factors mattered.
  • By 59 months, the development of a poor and under-nurtured child would lag by as much as 18 months behind her wealthier and more nurtured peers.
    • This child would not be able to recognize basic shapes like triangles or squares, count to 20, or understand temporal sequences.
    • She would also have gaps in her basic executive functioning and socio-emotional skills, including empathy and autonomy.
    • She would not likely be ready for school and may not have success once there.
  • Notably, however, if this same child, in the same poor household, were to have benefited from a nurturing environment, her level of development would have risen and would have started to approach levels found in children in wealthier but less nurturing households.
  • The nurturing environment thus appeared to mitigate the negative association lower levels of wealth had with the domains of development measured by the Engle scale.
Sources - Report(s) of results