Today's Chanakya's methods are unorthodox, but the media might not care.

On the day of the results, one polling agency joined the BJP and the AAP in being declared victorious – Today’s Chanakya. The media wrote flattering pieces and politicians tweeted their appreciation of the pollster. Yet not a lot is known of its research methods.

The political arm of the consumer research agency RNB Research, Today’s Chanakya has been “behind the scenes” for many years, but its work has been in the public domain for fifteen years. The agency had correctly predicted the last two general elections, the UK and US elections and various state elections, a Today’s Chanakya spokesperson said.

Today’s Chanakya’s political and electoral opinion polls are commissioned by political parties, individuals and the media, the spokesperson said. Its CEO V K Bajaj however declined to name any of its clients citing client confidentiality. He was not at liberty to discuss who had commissioned his much-lauded opinion polls for the five state assembly elections, either, Mr. Bajaj said. Unlike polling agencies like C Voter, AC Nielsen and the Centre for the Study of Developing (CSDS) whose electoral polls are exclusively commissioned and released by a media partner, Today’s Chanakya’s numbers are often put out by the agency through press releases or even on Twitter.

As for its research methods, Mr. Bajaj gave The Hindu the sample sizes for the five states – 4,595 people in Delhi, 11,280 people in Rajasthan, 13,890 people in Madhya Pradesh and 6,520 people in Chhattisgarh – but could not disclose the demographic composition of the sample, citing “respondent confidentiality”. “We have a database at the level of each assembly constituency of the age, income, caste etc of people, and we try to make our sample representative to it,” Mr. Bajaj said, but could not tell The Hindu what proportion of the sample was of a particular age or caste. While he could not disclose what proportion of assembly constituencies Today’s Chanakya polled in each state, each state was a “unique case”, he said.

Its polling methodology consists of face-to-face interviews, pen-and-paper questionnaires and “mystery shopping”, a tool usually used in market research that involves not disclosing to the respondent that he or she is being interviewed for a poll. Mr. Bajaj was not at liberty to disclose what proportion of the sample was polled using each method, he added.

As for its findings, Today’s Chanakya could only disclose the vote-share and the seat share numbers, Mr. Bajaj said, and not any data on the answers given by respondents because of “respondent confidentiality”, he said. “Vote share and seat share numbers are the only numbers we put out in the public domain,” Mr. Bajaj said, noting that Today’s Chanakya strictly complied with standards laid down by the market research industry body Esomar.

In the four state election results that brought the agency its recent successes, particularly in Delhi where the entry of a new party made forecasting complicated, the agency got the seat prediction closer than it did the vote-share, while for most other pollsters the reverse was true. While Mr. Bajaj said that political insight and understanding of a state was important for a pollster to keep in mind, he insisted that such subjective elements did not influence his predictions. “We rely solely on the data our field researchers come in with,” he insisted. Their algorithm of converting vote-shares into seats could not be replicated by another person, however, because of their expertise, Mr. Bajaj said.

Other pollsters did not want to comment on the record about a fellow polling agency and business rival, but said that the industry as a whole needed to move towards greater transparency about its clients and methodology. The media, which had demanded transparency in the run up to the polls, abandoned these ideals is favour of the “right” number, a pollster told me, asking not to be named.