By Perwinder Singh and Agastya Shukla
Public discussions on learning outcomes often become debates over numbers, with different assessment results presented as though one survey must prove the other wrong. However, differences between learning assessments such as the National Achievement Survey (NAS), the Annual Status of Education Report (ASER), and independent assessments conducted by organisations like Sanjhi Sikhiya (SS) do not necessarily indicate that one of them is inaccurate. Instead, these differences suggest that each assessment measures a different aspect of learning.
The more important question, therefore, is not which survey is “correct”, but whether our assessment system is capturing the knowledge and skills that children actually need.
To better understand this question, Sanjhi Sikhiya conducted a school-based assessment of Grade 3 students in government primary schools across four districts of Punjab during May 2025 and May 2026. Unlike NAS, however, the assessment focused not only on foundational reading and numeracy skills but also on whether children could understand what they read and apply basic mathematics in everyday situations. This allows the findings to complement, rather than replicate, existing national assessments.
According to the National Achievement Survey (NAS) 2024, nearly four out of five Grade 3 children in Punjab achieved the expected level in reading, and over three-quarters demonstrated basic numeracy skills. However, assessments that focus on foundational skills tell a different story. As shown in Table 1, both ASER and the Sanjhi Sikhiya assessments consistently found that substantially fewer children could read a simple story with understanding or apply basic mathematics in practical situations.
This contrast does not imply that one assessment is more accurate than another. Rather, it suggests that different assessments capture different dimensions of learning.
NAS consistently reports high reading and numeracy achievement in districts like Bathinda (80%), Fatehgarh Sahib (72%), Patiala (77%) and Ropar (81%). Sanjhi Sikhiya, however, shows far lower levels of story reading and comprehension viz. This application gap is exactly what current grade-level multiple-choice assessments find difficult to identify and without some independent checks, it stays invisible.
This is not a critique of NAS as an instrument. The problem arises when NAS is treated as a proxy for learning as opposed to a lens. Assessments are not neutral; they shape what systems prioritise. They inform the priorities in teaching, classroom practice and public accountability. What is taught becomes what is tested as widely recognised in education research, assessment systems influence what teachers prioritise in classrooms.
Less than 1 in 3 children can read a simple story, while just over half can do basic subtraction. On-ground assessment by Sanjhi Sikhiya in government primary schools reinforces the same gap.
As a result, a crucial question emerges: why such massive divergences in learning outcomes occur, in the same state, at the same grade level, at approximately the same point in time? The problem is not simply a question of data quality but of the metrics each survey seeks to gauge.
NAS is based on grade-level compliance. It gauges if children can answer questions based on the official curriculum they are supposed to have taken that year. It records coverage of syllabus and system delivery.
SS, on the other hand, poses a simpler question: can a child read, recognize, subtract and can a child count? It is intentionally simple in its tools, grounded more in foundational skills than grade expectations. Sanjhi Sikhiya takes it a step further to test not only the basic skills, but their application: whether the children can understand the things they read and whether they can apply maths in everyday life.
The Punjab puzzle becomes clearer when these tools are taken in unison. Punjab’s NAS results suggest that many children are meeting grade-level expectations. However, evidence from ASER and Sanjhi Sikhiya indicates that a significant proportion still struggle with foundational skills and their application.
While some variation between NAS and ASER is expected because they assess different aspects of learning, the magnitude of the difference is not the same across states. In Haryana, the gap in reading outcomes between the two assessments is only 2 percentage points, while Himachal Pradesh records a difference of 24 percentage points. Punjab, however, shows a much wider gap of 44 percentage points.
This does not imply that one assessment is more accurate than another. Rather, it raises an important question: why do different assessment tools present such contrasting pictures of learning in Punjab?
Punjab’s experience illustrates the consequences of failing to align assessment with actual learning. The divergence isn’t a statistical oddity. It is a warning signal.
For policymakers and practitioners alike, the implications are clear. The challenge is not to choose one of NAS or ASER, but to integrate them within a whole assessment ecosystem. Inclusion of foundational skills as a core curriculum, particularly by Grade 3’s conclusion, must be stressed explicitly.
The evidence presented through Tables 1 and 2 illustrates that different assessment tools (ASER/SS vs NAS) reveal different dimensions of learning. Rather than viewing these differences as contradictions, they should prompt deeper public discussion on how learning is measured and monitored.
The question isn’t whether Punjab’s learning challenge exists, but whether or not we’re prepared to see clearly when. Understanding this difference is the first step toward constructing an assessment system that does not just report, but facilitates progress.
This brings us to the deeper question of intent. At its core, the challenge is not merely one of measurement, but of intent. Before refining methodologies or expanding assessment frameworks, the state must first confront more fundamental questions about its priorities and commitments.
What proportion of the budget is actually being allocated to education, and more importantly, what outcomes is that spending producing?
If different assessment tools consistently tell different stories about learning, should the goal be to choose one over another—or to build an assessment system that brings these perspectives together?
Perwinder Singh is with Sanjhi Sikhya. Agastya Shukla is Program Associate at PANJ Foundation.