Voice Personas
Three hundred field researchers are three hundred slightly different instruments. Each has a manner, an accent and a way of nudging, and the correlation that induces within one enumerator's workload inflates variance in a way nobody measures. At a modest intra-interviewer correlation of 0.02 and 120 interviews each, a nominal 3.5 lakh field sample carries the precision of roughly 1.03 lakh.
Personas here do the opposite. Assignment is randomised within stratum, not matched to maximise cooperation. Matching would raise the response rate and quietly confound the result. Randomising means the persona effect becomes a measured quantity that can be corrected for, which is the table at the bottom of this screen.
Language and register matching is different and uncontroversial: a Purvanchal register for eastern Uttar Pradesh raises comprehension without touching who is willing to answer.
| Persona | Assigned | Cooperation | Q11 refusal | Effect on estimate | Correction |
|---|---|---|---|---|---|
| Asha Hindi | 16.2% | 46.2% | 13.8% | +0.3 pts | Applied |
| Ramesh Hindi | 14.8% | 49.1% | 11.2% | −0.4 pts | Applied |
| Devendra Hindi | 3.0% | 47.8% | 12.4% | +0.1 pts | Applied |
| Nusrat Urdu | 6.4% | 44.7% | 16.4% | +0.9 pts | Applied |
| Ananya Bengali | 9.1% | 45.9% | 14.6% | −0.3 pts | Applied |
| Sneha Marathi | 8.6% | 43.8% | 14.1% | −0.2 pts | Applied |
| Gopal Telugu | 7.9% | 47.3% | 12.6% | −0.5 pts | Applied |
| Karthik Tamil | 7.2% | 41.6% | 18.9% | +0.6 pts | Applied |
| Bhavana Kannada | 6.1% | 44.2% | 15.3% | +0.2 pts | Applied |
| Jignesh Gujarati | 5.8% | 48.4% | 11.9% | −0.1 pts | Applied |
| Sasmita Odia | 4.4% | 46.7% | 13.1% | +0.4 pts | Applied |
| Meera Malayalam | 4.2% | 42.9% | 17.2% | +0.7 pts | Applied |
| Harleen Punjabi | 3.9% | 45.1% | 14.8% | −0.2 pts | Applied |
| Kalpana Assamese | 2.4% | 44.8% | 13.9% | +0.3 pts | Applied |
Effect is the difference in the incumbent vote share estimate against the persona-pooled mean, after stratum controls. Randomisation is what makes this column interpretable at all.
One interviewer, two lakh calls, identical wording and pacing every time. There is no interviewer variance term because there is one interviewer. That is arithmetic, and it is the hardest thing for a field-based competitor to answer.
No persona is assigned on inferred community, religion or caste. Voice, name and register are selected on language and region only. Inferring community from a name and then choosing a matching voice is a short walk to a product we will not build.