Delhi municipal councillors care more for performance ratings in newspapers: Study

Scholars from American universities informed some municipal councillors in Delhi that a newspaper would report on their performance just before elections. This is what happened.

Representational image | People reading newspapers at the central park of Connaught Place in New Delhi | Photo: T. Narayan | Bloomberg

Why are policy choices in lower-income democracies often unresponsive to the development preferences of their electorates? Is it because politicians expect voters to remain ignorant of their performance record? Or is it because the votes of poor citizens, even when they are informed and constitute a majority, are largely determined by ethnic loyalties and vote-buying ploys?

We use a set of field experiments, conducted at scale in Delhi, one of the world’s most populous cities, to demonstrate that a credible promise of revealing performance information to voters makes politicians more responsive to voter preferences. It also makes voters’ electoral choices more responsive to politicians’ performance. Finally, it alters the pool of politicians: parties start weeding out those with the weakest records.

The setting

Our field experiments occurred in the context of Delhi’s Municipal Corporation, an elected local body. A councillor enjoys multiple policy levers. She receives an annual discretionary development fund to spend on local development, typically improvement of local infrastructure.

Our primary intervention was conducted at scale, covering 240 of Delhi’s 272 wards. In each ward, citizens elect a single member for a five-year term. In 2010, two years prior to Delhi’s municipal elections, a random sample of incumbent councillors were informed that a leading Hindi newspaper would report on their performance a month before the next election.

For a randomly chosen subset of treated councillors, the newspaper also published midterm report cards, identical in structure to the final report cards, in May–June 2010. This potentially enhanced the credibility of our primary intervention.

Prior to the 2010 information campaign, we surveyed slum-dwellers in over 100 high-slum density wards on their spending preferences in order to obtain preference weights for different items in the infrastructure budget.

We document a large divergence between slum dwellers’ infrastructure priorities and councillors’ spending choices. Nearly 70 per cent of our respondents describe sanitation (sewage and drainage) as a concern, but only 16 per cent of councillor funds are spent on it, against 54 per cent on roads, despite the fact that only 2 per cent of our respondents complain about roads. Given this disparity, we hypothesise that at least for high-slum wards, defined as wards where the fraction of slum area is above the median value computed across all wards (45 per cent), anticipated public disclosures of spending patterns should incentivise councillors to move spending in a more pro-poor direction.

The method

To directly evaluate the effect of providing the councillors information about the needs of their constituents, we implemented a second cross-cutting experiment within the set of high-slum wards.

Motivated by the central importance of sanitation problems in the slums, we provided a random sample of councillors with State of Sanitation Information (SSI) for their wards. This was based on our own audits, and included geo-located information on the quality of public toilets, sewers, and garbage removal in three slums. The information was collected and disseminated twice – eight months and two months prior to the election. The information was never made public, and allows us to ask whether informing councillors about sanitation issues – something the report cards might have done as well – influences their performance.

The findings

We have three sets of findings.

First, anticipated public performance disclosures led councillors in high-slum wards to make infrastructure investment decisions that were 0.62 standard deviations more pro-poor. We also observe a 0.37 standard deviation increase in assembly and committee attendance in this group.

The estimated treatment effects are similar across councillors who only anticipated disclosure in 2012 and those who additionally received a midterm report card in 2010. This argues against any additional value from early information associated with the 2010 report card.

Second, consistent with the absence of a salience or information effect from the midterm report cards, but even more striking, actionable information provided via the private SSI intervention did not improve sanitation in slums. Why would councillors react to the newspaper report cards in order to please the voters, but not to the audits?

One possibility is that the set of voters who directly benefit from the sanitation improvements (the slum residents) is too small to be politically relevant – perhaps because there are no credible mechanisms by which this performance information would reach a broader voting public. The performance report cards, in contrast, have an amplified effect due to publication in a newspaper.

The alternative possibility is that voters put little weight on policies and, instead, focus on the candidate’s ethnic identity or money provided by parties on the election eve. In this case, newspaper report cards matter because it provides party selection committees with performance information which they use to allocate party tickets.

To evaluate these explanations, we turn in our third set of findings on how the treatment impacts party ticket allocation and, subsequently, voter behaviour.

In the 2012 elections, ticket allocation gained greater salience due to an unanticipated expansion in gender quotas for councillor seats. In January 2012, the government announced that the number of wards reserved for women in the April 2012 elections would increase from 33% to 50%, with the choice of reserved wards being randomized. This created a randomly chosen set of 80 incumbents (from our sample of 240 incumbents) who were ineligible for election in their current ward (now on ineligible councillors).

Their parties had to decide whether to allocate them party tickets to other (unreserved) wards.

Media disclosures influenced party ticket allocation. Treated incumbents were 12 percentage points more likely to run for re-election. The treatment effect captured the ticket allocation to ineligible councillors with a more pro-poor spending record. Specifically, ineligible treated incumbents with a pro-poor spending a standard deviation above their counterparts were 13 percentage points more likely to get a party ticket for a different ward.

These ticket allocation effects translate into a higher vote share. (Naturally one has to run to gain votes, but running does not guarantee winning). Treated ineligible incumbents with a pro-poor spending record that is one standard deviation above their counterparts are 23 percentage points more likely to win in the next election.

Newspaper effect

Newspaper report cards may have independent advertisement effects over and above any information effect. For one, publication in a newspaper makes the information more credible. Even the voters who probably did not have the full information may be subject to an advertisement effect – they may react to the fact, for example, that the information about the candidates (that they perhaps already knew) is now common knowledge.

That publication in the media plays such a critical role in decisions at all levels is important at a time when media freedom is threatened. It is also worrying, since it suggests that politicians and even parties may not care about performance unless it gets highlighted in the media. Since space in the credible print media is limited, this may limit the scope for providing effective incentives to politicians.

This suggests an important direction for future research: is there some way to provide the public with credible performance information on a routine basis, and get them to pay attention to it, without actually printing report cards in the newspaper before the election?

This is an edited excerpt from the paper titled ‘Public Information is an Incentive for Politicians: Experimental Evidence from Delhi Elections’ by Abhijit Banerjee, Nils Enevoldsen, Rohini Pande and Michael Walton. Authors are affiliated with MIT (Banerjee), Yale (Pande), Harvard (Walton), and unaffiliated (Enevoldsen). 

Views are personal.