Permission, Not Power Why Pakistan’s Work-From-Home Experiment Was Constrained by Administration, Not Infrastructure

Summary

This Policy Viewpoint evaluates Pakistan’s March–June 2026 work-from-home experiment using a survey of 541 federal employees across 21 departments. It asks a simple question: if WFH is used again, what must change for it to work better?

The main constraint was institutional readiness, not household readiness. Most employees had the basic technology and power needed to work from home, but fewer than half had remote access to official systems or clear guidance, and only two in five could reliably access required files. Output was broadly maintained, but working hours fell. Supervisors often reported weaker team performance, and official work shifted heavily to WhatsApp. Fuel savings were real, but frequent recalls to the office reduced them.

For any future WFH arrangement, the government should:

• provide secure remote access to official systems and files;
• set clear work, reporting, and performance rules;
• move official communication to secure, auditable platforms;
• protect designated WFH days from routine office recalls; and
• apply WFH selectively to functions that can be performed remotely.

The central lesson is simple: treat WFH as an administrative reform, not just a change in work location.

1. The policy, and why it is worth evaluating

On 9 March 2026, the federal government notified a four-day working week and directed that up to half of official staff work from home, as part of a fuel-conservation and austerity package prompted by the regional energy shock. The federal week was compressed to Monday–Thursday, banking, industry and agriculture were exempted, and departmental oil allowances were halved, among other measures (Aslam and Hussein, 2026; Zulfiqar, 2026). The measure ended in June.

The episode is analytically valuable for three reasons. First, almost all credible evidence on remote work comes from voluntary arrangements in private firms in high-income countries (Bloom et al., 2015; Barrero, Bloom and Davis, 2023; Bloom et al., 2024). This was compulsory, government-wide, and it took place in a lower-middle-income bureaucracy. Second, the developing-country evidence that does exist is drawn from firms with quantifiable output and finds real losses; i.e., an Indian data-entry trial found that workers randomly assigned to home were 18% less productive (Atkin et al., 2023), and an Indian IT firm recorded longer hours with lower output (Gibbs et al., 2023). Neither speaks to a bureaucracy, where the product is a decision on a file rather than a keystroke. Third, the policy was justified on fuel savings, a claim the energy literature treats with caution because rebound effects routinely erode headline reductions (Hook et al., 2020; O’Brien and Aliabadi, 2020).

What was missing was evidence on whether a state that runs on physical files, biometric attendance and face-to-face approval can disperse and still function. This study was commissioned to answer that question.

2. Data and approach

The survey, fielded after offices fully reopened, covers 541 consenting federal employees from 21 ministries, divisions, attached departments, and autonomous bodies in Islamabad. Of these, 56.9% personally worked from home during the policy, and 46.6% supervise staff, allowing every productivity claim an employee makes about themselves to be set against a supervisor’s judgement of a team. Respondents span the grade structure: 42.5% at BPS 11–16, 41.2% at BPS 17–18, 12.0% at BPS 19 and above. Some 20.1% of respondents are women.

Because the policy had ended, the instrument was retrospective by design; respondents recalled the period when the rotation was in force and compared it with how they normally worked in the office. Indices summarise the survey’s structural findings, each built from binary items scored 0 or 1 and averaged onto a 0–100 scale, so that the components can be read directly and audited. Throughout, the analysis separates what the employee already had in place from what the department supplied, a distinction that drives the paper’s main result.

3. Implementation broadly held, and coordination did not break

The first result is that implementation held. Some 60.8% of respondents were in offices running the notified four-day week with a rotation, and only 6.8% remained on a five-day week with full attendance. Where a rotation operated, some offices (82.8%) ran it by splitting staff into two fixed groups on fixed weekdays, 68.6% by setting days through a written roster issued by the administration, and 72.6% by assigning two days at home.

That design choice mattered. Fixed, synchronised cohorts are precisely what the coordination literature recommends, as they preserve overlapping hours and prevent the asynchronous dispersion that degrades team performance (van der Lippe and Lippényi, 2020). The results support this: 75.8% of home workers found it easy or very easy to coordinate with colleagues; 73.0% reported that the colleagues they needed were usually or always available during the same hours; and only 4.3% cited colleague scheduling as an obstacle. This is the study’s clearest positive finding, and it disposes of the most common administrative objection to rotation, namely that the office would simply stop functioning. It did not.

4. A quarter of staff have no remotable work at all

Feasibility is not uniform, yet the policy was applied as though it were. Asked what share of their tasks could be done from home, 23.0% of federal respondents said none, and 43.6% said a fifth or less; only 26.4% said more than 16%. This is broadly consistent with national estimates of Pakistan’s remotable job share (Hasan et al., 2021) and with Dingel and Neiman’s (2020) occupational approach. 

Table 1 Take-up and reported relative output, by task remotability

         Source: Authors, Work from Home Survey, 2026

Take-up[1] tracked remotability, rising from 12.8% among those who say none of their work can be done at home to 83.3% at the top of the scale, suggesting supervisors exercised some judgement. But the residual is the problem, as staff with no remotable work were still sent home, and a quarter reported doing much less, compared with 1.9% among the most remotable. Take-up also rose with seniority 50.0% at BPS 11–16, 60.3% at BPS 17–18 and 72.6% at BPS 19 and above, a gradient with no obvious basis in task content, which points to allocation by discretion rather than by rule.

1. Output held on average; hours did not, and supervisors dissented

For the measure with an explicit pre-policy reference, weekly output was largely unchanged: 15.9% produced more and 14.8% produced less. Quality and timeliness were similarly stable, with 75.0% reporting quality as good as before and 69.5% finding deadlines no harder to meet. Pending workload changed little, with 74.0% reporting no change in backlog. 

Table 2 Output and supervisor reports

 Source: Authors, Work from Home Survey, 2026

Two qualifications cut against a benign reading. The first is that supervisors judging their own teams are markedly more negative: 39.0% said team output fell, compared with 14.0% who said it rose. Self-assessment and supervisory assessment diverge, and international evidence suggests the supervisory view deserves weight, since self-reported effort under remote work is systematically flattering (Gibbs et al., 2023).

The second is hours, which shows the sharpest result in the survey. Home workers recorded a mean of 8.2 hours on an office day versus 5.9 on a home day, a 27.8% decline. On an office day, 85.4% worked eight hours or more; on a home day, only 33.1% did, and 18.2% worked three hours or less.