LOST SLEEP, LOST STUDY TIME: UNEQUAL ELECTRICITY OUTAGES AND THEIR DISRUPTION OF LOW-INCOME AND HIGH-INCOME HOUSEHOLDS IN KARACHI

Abstract

In Karachi, K-Electric (KE) schedules load-shedding by feeder, assigning longer outages to feeders with high aggregate technical and commercial (ATC) losses. This design links a household’s outage burden to the behaviour of its neighbours, and because high-loss feeders cluster in low-income and informal areas, it raises an equity question. This paper reports a pilot household survey of 52 households in three Karachi areas, collected in 2026: Lyari (image), Orangi Town (image) and DHA (image). Households reported outage frequency and duration, the longest outage of the week, the outage pattern, backup equipment and its running cost, and the biggest disruption caused. Lyari households reported a mean of 7.2 outage-hours per day (bootstrap 95% CI 5.1 to 9.4), Orangi Town 2.3 (1.4 to 3.3) and DHA 0.08 (0.0 to 0.2). Differences across the three areas were large and statistically significant (Kruskal–Wallis image), and the contrast between the two lower-income areas and DHA corresponds to a rank-biserial effect size of about 0.79. Median monthly electricity bills in DHA were roughly nine times those in Lyari and Orangi Town combined. Every DHA household had a UPS or generator, while 2 of 16 Lyari and 5 of 18 Orangi Town households did; no katchi abadi household reported any backup (0 of 14). About 89% of all reported outage-hours were borne by households with no backup at all. Among the few lower-income households that ran generators, the generator’s running cost equalled a median of about 56% of their electricity bill. These patterns suggest a two-layer inequity: some people face more outages than others, and not everyone has the same ability to protect themselves from these disruptions. Since the sample size is small and not randomly selected, it’s important to view these results as preliminary findings that could pave the way for a more extensive study using detailed feeder-level data.