Eighteen Missing Households at Tola-B: Sample Replacement After River Erosion
You are a field assistant attached to a block statistical office in Saharsa district, Bihar. Your sample list contains 120 households spread over four tolas of one gram panchayat, and the survey must be finished in three days. Your supervisor sits 40 km away and is reachable only by mobile phone. On the second morning, at Tola-B, you find that 18 listed households no longer exist: 11 families shifted to a relief colony after the Kosi embankment breach two seasons ago, and the remaining 7 houses stood where the river bank has since eroded. The ward map you carry is a photocopy from the 2011 census. The local anganwadi worker knows most families and offers to point out where the shifted ones now live. A ward member suggests you simply fill the gap with nearby houses so that the total of 120 is met and the team can leave on time.
1. What exactly do you record for the 18 non-existent households, and why is the ward member's substitution suggestion unacceptable?
Read the worked answer
Record the status; do not replace on the spot.
For each of the 18 listings: mark the household as not found / non-existent with a separate reason code for 'shifted after flood' (11 cases) and 'structure lost to erosion' (7 cases). Take a GPS point at the listed location, not at any new one, photograph the eroded bank, and note how far the bank has moved.
Substitution as suggested is unacceptable because the sample was drawn with known selection probabilities. Hand-picking nearby houses turns a probability sample into a convenience sample: the houses easiest to reach are usually roadside and better-off, so the 120-household estimate would be biased and no longer comparable with other tolas. If the design permits replacement, substitutes must come from the reserve list drawn in advance for the same tola, taken in the listed order, and only after the supervisor's approval.
Sequence: record status → report → obtain approval for reserves → survey reserves → note both the original and the substitute ID in the remark field.
2. How do you use the anganwadi worker's local knowledge and the 2011 ward map without corrupting the sample design?
Read the worked answer
Treat both as aids to locating, never as aids to selecting.
The anganwadi worker can guide you to the relief colony so you can attempt the schedule with the shifted families at their new location. If a household is traced and interviewed, keep the same household ID, add the new GPS and write 'shifted, same household, traced through anganwadi worker'. She must not decide which families are 'similar enough' or add names to the list - that would be selection by a local informant and would break the sampling rule.
Use the 2011 map only for the tola boundary and the position of listed structures. State on the form that the map is of 2011 vintage and that part of the tola has since moved. If an entire tola has relocated, that is a listing problem for the supervisor, not something to fix quietly in the field.
Both sources are evidence, not authority: your record must separate what you observed (GPS, erosion, empty site) from what you were told (families now in the colony).
3. You have two minutes of phone time. Write the report you give your supervisor, in priority order.
Read the worked answer
1. Finding: At Tola-B, 18 of the 120 listed households are non-existent - 11 shifted after the Kosi breach, 7 lost to bank erosion. GPS taken at each listed point; forms marked with reason codes.
2. Decision needed today: May I use the pre-drawn reserve list for Tola-B, in listed order, to fill these 18? I will not substitute on the ward member's suggestion without your approval.
3. Design issue: Part of the tola has shifted as a group; the ward map is from 2011. A revised listing may be needed for this panchayat, not just replacements.
4. Plan meanwhile: I will complete Tola-A, C and D today and return to Tola-B tomorrow morning to trace the shifted families in the relief colony with the anganwadi worker, keeping their original IDs.
5. Request: Confirm by message so I have your approval in writing for the file.