Does Coverage Follow the AQI? A Constructed-Week Content Analysis of Air-Pollution Reporting in Two Indian Dailies

Does Coverage Follow the AQI? is a BJMC Semester 6 Media Project that asks a simple question with a measurable answer: does newspaper coverage of air pollution rise and fall with the air itself, and do English and Hindi readers get the same story?

Two members analyse air-pollution coverage in two real Indian dailies — Hindustan Times (Delhi edition, English) and Dainik Jagran (Delhi edition, Hindi) — across calendar year 2025. Checking all 730 editions is impossible for a student team, so the sample uses the constructed-week method: within each half of the year, one date is drawn at random for each of the seven weekdays, giving two constructed weeks per newspaper — 28 editions in all. Every pollution-related item in those editions (report, feature, editorial, opinion, letter to the editor) is coded on a written codebook: date, page, prominence, word count, genre, dominant frame (health, governance, politics, science, civic action or solutions), source types quoted, whether official AQI or GRAP data is cited, and visuals.

Coverage volume per quarter is then placed beside the CPCB's published AQI record for Delhi to test whether coverage tracks the air, and framing is compared between the two papers with chi-square. Both members independently code 15% of items so intercoder reliability can be reported honestly.

The bundle contains the full design, a draft codebook, the reliability and analysis plan, a report scheme for the 30 CCA + 70 ESE evaluation, a 12-slide presentation outline and a viva bank.

Air pollution is Delhi-NCR's most repeated public-health emergency, and newspapers are the record of how citizens and governments talked about it. But claims about that coverage — that it spikes only in November, that it blames farmers and governments but ignores solutions, that Hindi readers get thinner and more event-driven stories than English readers — circulate as impressions, not measurements. No sampled, reliability-tested description of the beat exists that a journalism student can point to.

There is a need for a systematic content analysis built on a defensible probability sample (the constructed week), a written codebook with tested intercoder reliability, and a benchmark — the CPCB's own AQI record — against which the rhythm of coverage can be judged. This project provides that measurement for two Delhi dailies across 2025 and turns the result into practical recommendations for environment reporting.

Abstract

This study uses quantitative content analysis to examine how two Delhi dailies — one English (Hindustan Times), one Hindi (Dainik Jagran) — covered ambient air pollution across 2025. A constructed-week sample (one randomly drawn date per weekday, per half-year, per newspaper) of 28 editions yields all pollution-related items, which are coded for prominence, genre, frames, sources, data citation and visuals. The study tests whether the volume of coverage tracks the quarterly air-quality record published by the CPCB, and whether frame and source hierarchies differ between the English and the Hindi paper.

Introduction

Delhi's air is a seasonal emergency: every October to January the AQI crosses "severe", the Graded Response Action Plan (GRAP) kicks in, construction stops and schools close. Newspapers respond with front pages, explainers and editorials. Yet coverage in March, when the air improves, has barely been described — and the question of whether newspapers follow the air or the news calendar is rarely answered with data. Whether Hindi-language readers, who form the larger part of the NCR audience, receive the same framing as English readers is likewise an open, answerable question.

Literature gap

Textbook treatments of content analysis (Wimmer and Dominick; Krippendorff) and the constructed-week sampling research of Riffe, Aust and Lacy are well established, and Indian media research has examined election and health coverage. But systematic, student-replicable studies of Indian environment-beat coverage that (a) use a defensible probability sample and (b) link content volume to actual monitored air-quality data remain scarce. Most available writing on Delhi's pollution media coverage is commentary, not measurement.

Existing practice vs this study

• Existing: anecdotal claims that "the media wakes up only in November"; one-off critiques of single front pages; no sampled counts; no reliability testing; no comparison across languages.
• Proposed: a documented, reproducible constructed-week sample; a piloted codebook; intercoder reliability on 15% of items; statistical comparison of coverage volume against CPCB AQI data and of frames between the two papers.

Feasibility

• Technical: both e-paper archives are accessible through the department's library subscription or short-term personal subscriptions; coding runs on Excel; the statistics are chi-square and one rank correlation — well within BJ-403's scope.
• Operational: screening 28 editions and coding an estimated 150–250 items is realistic for two coders in about five weeks.
• Economic: under ₹1,000 for archive access; no production costs.
• Ethical: published newspaper content is analysed with brief quotations and full attribution only; no private individuals are studied and no personal data is collected.

In scope

• Hindustan Times (Delhi edition) and Dainik Jagran (Delhi edition), 1 January – 31 December 2025, print editions only.
• All news items in the sampled editions about ambient air pollution: reports, features, editorials, opinion columns and letters.
• Comparison of coverage volume with the CPCB AQI record for Delhi.
• Frame and source analysis across the two papers.

Out of scope

• Television, radio and online-only content of the same publishers.
• Other environmental beats (water, noise, waste) and other cities.
• Audience reception — this study measures content, not readers.
• Items outside the randomly drawn editions, however tempting; no substitutions beyond the documented fallback rule.
• Paid content, advertorials and sponsored supplements.

Design

A quantitative content analysis with a constructed-week probability sample, following the sampling logic tested by Riffe, Aust and Lacy (1993) and the textbook procedure in Wimmer and Dominick. The unit of analysis is the individual news item; the sampling unit is the edition.

Sampling (weeks 1–3)

1. Divide 2025 into two halves: January–June and July–December.
2. For each half, for each newspaper, draw one date at random for every weekday (Monday to Sunday) using Excel's RAND() against a numbered list of dates; record every draw and its seed in a sampling log so the sample can be regenerated exactly.
3. Fallback rule: if a drawn edition is missing from the archive (holiday, page unavailable), take the nearest previous edition and log the substitution. No other substitutions are allowed.
4. Expected sample: 14 editions per newspaper, 28 in total.

Item screening (weeks 4–5)

Within each sampled edition, screen for pollution items using a fixed list of search terms — pollution, air quality, AQI, PM2.5, smog, GRAP, stubble, pradushan, hawa ki quality, dhundh — applied to the e-paper's text layer, then a visual skim of every page (captions and headlines do not always contain the terms). Screen decisions are logged per edition.

Codebook and pilot (weeks 5–6)

Draft codebook variables: date and paper; page number and page type (front page / city / editorial-opinion / other); item genre (report, brief, feature, editorial, opinion, letter); word count band; dominant frame — health, governance and enforcement, politics and blame, science and data, civic action and solutions, other; source types quoted (up to three): government official, doctor or health expert, scientist or CPCB/SAFAR data, citizen, farmer, activist, politician, none; explicit AQI or GRAP figure cited (yes/no); visuals (photo, infographic, none). Pilot on two non-sample editions, revise wording, freeze the codebook.

Coding and reliability (weeks 6–9)

Both coders independently code a 15% reliability subsample (minimum 25 items). Compute Holsti's percentage agreement and, where table software allows, Krippendorff's alpha; target agreement ≥ 0.80 — disagreements are adjudicated by discussion and the codebook clarified before the remaining items are split for full coding.

Analysis plan (weeks 10–12)

• Descriptives: items per edition, page placement, genre mix, frame distribution, source frequencies.
• Coverage vs air: quarterly items-per-edition rate against the CPCB's quarterly mean AQI for Delhi — Spearman's rank correlation.
• English vs Hindi: frame distribution and source hierarchy compared with chi-square (expected counts below 5 are merged into adjacent categories and the merger is reported).
• A short qualitative read of the ten highest-prominence items contextualises the numbers.

Timeline

The study is a measurement pipeline: a probability sample feeds a coded dataset, which is joined to monitored air-quality data and analysed. The figure below belongs in the Methodology chapter.

flowchart TD
  A["Define population: all 2025 editions, both dailies"] --> B["Split into two half-years"]
  B --> C["Random draw: one date per weekday per half (documented in sampling log)"]
  C --> D["Retrieve 28 sampled editions from e-paper archives"]
  D --> E["Screen with fixed search terms and page skim"]
  E --> F["Item list: reports, features, editorials, letters"]
  F --> G["Pilot codebook on two non-sample editions"]
  G --> H["Freeze codebook"]
  H --> I["Double-code 15% reliability subsample"]
  I --> J{"Agreement >= 0.80?"}
  J -->|No: revise definitions, recode| I
  J -->|Yes| K["Split full coding between coders"]
  K --> L["Coded dataset in Excel"]
  M["CPCB quarterly AQI data for Delhi"] --> N["Join to coverage counts"]
  L --> N
  N --> O["Spearman correlation: coverage vs AQI"]
  L --> P["Chi-square: frames and sources by paper"]
  O --> Q["Findings, recommendations, report"]
  P --> Q

Codebook at a glance

Every coded item also keeps its exact citation (paper, date, page, headline) so any finding can be traced back to the original page — the content-analysis equivalent of auditability.

• Add television bulletins and the papers' websites for the same dates to compare platforms on one beat.
• Repeat over three years to see whether the coverage rhythm is stable or event-driven (a multi-year constructed-week design straight from Riffe, Aust and Lacy).
• Add Lucknow or Kanpur editions to test whether the pollution story is treated as Delhi's alone or a Gangetic-basin story.
• A reception arm: brief interviews with desk editors on why the beat collapses in March.
• An audience-side experiment: showing sampled headlines to readers to measure perceived severity against the AQI they experienced.