Despite BJP’s demonetisation policy having some negative economic impacts in India, why did they win the Uttar Pradesh state election?
Experiments in macroeconomic policy are rare. One much discussed exception is India’s recent demonetisation. On 8 November 2016, all currency notes worth 500 rupees (~$7.50) and 1,000 rupees were made invalid overnight. These were the two highest-denomination notes and constituted nearly 86% of the country’s cash. People holding cash were given until the end of the year to deposit it into their bank accounts.
The following process of reissuing new currency in denominations of 500 and 2,000 rupees was slowed by numerous implementation issues. For instance, the new 2,000 rupee notes did not fit existing ATM machines (Tharoor 2016). There was also a large deficit in the total amount of cash ready for circulation — even on 23 December 2016, month and a half after the demonetisation was announced, the currency in circulation was only about half of the pre-demonetisation amount (Mazumdar 2016). Consequently, banks were forced to impose severe limits on cash withdrawals. People were allowed to withdraw no more than 2,500 rupees (~$40) on any given day from ATMs, and the total weekly cash withdrawal in any form (ATM, bank tellers, etc.) was capped at 24,000 rupees (~$370).
Conflicting economic and political views
With some prominent exceptions, economists across the ideological spectrum were broadly negative about the move as an economic policy. First, there was the potential cost of a massive liquidity crunch, in which the volume or number of economic transactions reduces due to insufficient cash holdings. The brunt of this cost was borne by the informal sector, where 85% or more of the Indian labour force is employed, as transaction here have been traditionally carried out in cash. Second, one of the stated motivations of the policy was to reduce corruption, yet the value of the highest-denomination bill was doubled, making it easier to pay people illegally with anonymity. Thus, demonetisation seemed more like a one-off penalty on those who were holding large quantities of cash at the time without affecting the future incentives for corruption.
However, many analysts in the press described this policy as Prime Minister Modi’s master move. They felt that it affirmed his commitment to fight corruption and his willingness to be tough when needed. Some predicted it would help him electorally. This prediction was confirmed when the political party which he belongs to, the Bharatiya Janata Party (BJP), won a substantial majority of seats (77%) in the March 2017 state-level elections for India’s most populous state, Uttar Pradesh. Though receiving only 40% of the votes, this was a 25 percentage points increase in vote share as compared to the 2012 assembly elections.
Where did the economists’ analysis go wrong? Were the economic consequences less negative than expected? Did people approve of the policy despite being hurt, and if so why? Lastly, did that approval ultimately turn into votes for Mr. Modi’s party?
Economic consequences in Bangalore: 20% fall in sales
To assess the economic consequences, we combine primary survey data from Bangalore, Karnataka, with secondary data on wholesale markets. We surveyed 400 wholesale and retail traders in Bangalore in December 2016 and early January 2017 (201 wholesalers, 199 retailers).1 We asked them how the demonetisation impacted their sales. While no change was reported by 29% of respondents, 20% of them reported a fall in sales of more than 40%. Further, sales were about 20% lower on average. These impacts are of course self-reported and potentially self-serving. Hence, we compare this estimate to one generated with sales data from two markets in Bangalore that had consistent data between 2011 and 2016. We use sales data from 2011 to 2015 to predict the sales for November and December 2016 using panel data methods and cumulative rainfall to predict sales.2 We find the ratio of actual to predicted sales is about 0.83, which is similar to the reported impact of about 20% lower sales.
The view of market participants: 73% felt that demonetisation was good for the country
What were traders’ stated reactions to this large economic dislocation? In striking contrast to the size of their actual sales losses, 73% of the sample agreed with the statement “It is a good thing for the country and the economy and it is only a slight inconvenience for me”, while only 6% expressed the view that it was bad for the country.
Why would the traders have such a positive view of what appears to be a massive negative shock to their own returns? One possibility is that they read it as a signal of Mr. Modi’s intentions, thereby making them feel better about the prospects of the country. A second and less charitable interpretation is that they were glad that people with large amounts of cash (the ‘corrupt rich’) suffered losses due to the policy.
A potentially revealing reason emerges from another survey carried out in rural Odisha in which respondents were asked how much they thought different parts of society were hurt by the demonetisation (Banerjee et al., in progress). Since the survey had only 70 respondents, it is at best suggestive though nonetheless interesting with 88% of the respondents saying that the demonetisation was good for them. This result comes despite their response that all groups were badly hurt economically. The median response for all groups on how much they were hurt was four or five on a scale of one to five, with five being the most damage. Perhaps the pleasure from thinking about corrupt politicians suffering losses explains this response as those surveyed felt that politicians, ahead of farmers and workers, were most negatively affected by the demonetisation.
Ironically, it does not appear that the ‘corrupt rich’ lost too much. News reports suggested that 90% or more of the stock of demonetised currency came back to the banking system, suggesting that the owners of cash found a work-around, potentially through corrupt means.3 But it is not clear that the man on the street came to understand this; the government actively intervened to cut off information on how much cash was flowing back in.
Impact on Uttar Pradesh: economic and political
Perhaps we are over-interpreting people’s reported opinions. To see how voters actually reacted to the demonetisation, we focus on the Uttar Pradesh elections. We first establish the enagtive economics impact of demonitisation in Uttar Pradesh. For the ‘mandis’ (wholesale markets) in Uttar Pradesh, we constructed a data set of daily wholesale prices, quantities brought to the mandis, and sales for a range of crops from 2011 to 2016. There are 131 such mandis in Uttar Pradesh that reported data both before and after demonetisation.4 We then matched each mandi to rainfall data from the Tropical Rainfall Measuring Mission. Next, we created a predictive model of the sales of specific crops in the mandi in normal circumstances (i.e. without demonetisation) in 2016, given the rainfall in 2016 around that mandi. The average proportion of actual to predicted sales using this model is about 0.77, meaning that actual sales were about 77% of predicted sales,5 but it varies substantially across mandis.
We now analyse the impact of shortfall in sales on voting behaviour by matching the sale shortfall in the mandi closest to a constituency on the vote share in that constituency for the BJP. The assumption here is that the sales shortfall was the result, direct or indirect, of demonetisation and that voters knew this. To estimate the effect of this shortfall on voting behaviour around each mandi, we examine how the vote shares of the BJP vary as a function of the gap between the actual and predicted sales in the local mandi. We find that — controlling for predicted sales, distance to the mandi, and the total number of candidates in the constituency election — a 100% decrease in sales reduced the BJP vote share by about 0.0045 percentage points (significant at the 5% significance level). This means that a 100% decrease in sales at the mandi closest to the constituency resulted in a 1% fall in the BJP mean vote share in that constituency, which is quite a small effect, although in the expected direction
This result is far from definitive. The BJP may have been blamed for other problems that affect sales but are unrelated to the demonetisation. Moreover, we are at best only picking up the relative difference in vote shares across locations affected to different degrees. The demonetisation might have boosted the BJP’s overall vote share, at least among those unaffected by the fall in sales and those benefitting from lower prices. However, there is some slight solace here for those who prefer to think that bad economics does not go entirely unpunished.
Banerjee, A, E Breza, A Chandrasekar and B Golub, “Combating Rumors: Evidence from a Field Experiment During the Indian Demonetization”, work in progress.
Mazumdar, S (2016), “While the RBI Is Silent, Its Numbers Tell Us Demonetisation Has Failed”, The Wire, 29 December.
Tharoor, S (2016), “India’s Demonetization Disaster”, Project Syndicate, 6 December.
 We approached sellers in these markets and asked them to take the survey. The response rate for retailers was 66.67% and for wholesalers was 33.3%.
 We use market-by-year fixed effects and year-by-month fixed effects here.
 See http://timesofindia.indiatimes.com/india/90-of-scrapped-notes-back-in-system-big-dividend-unlikely/articleshow/56210235.cms, https://www.forbes.com/sites/timworstall/2016/12/28/90-of-scrapped-notes-in-banking-system-india-gets-the-tax-not-the-dividend/#349d28672eae, and http://indianexpress.com/article/business/economy/demonetisation-black-money-cash-crunch-rs-14-lakh-crore-in-old-notes-are-back-only-rs-75000-crore-out-rbi4465542/
 Not all markets report prices consistently.
 We use market-by-year fixed effects, year-by-month fixed effects, and cumulative rainfall to predict sales. An alternate predictive model that replaces market-by-year fixed effects with interactions of market fixed effects with linear, quadratic, and cubic monsoon rainfall in the above specification has an average proportion of about 0.82.
Photo credit: Credit: jayk7/Getty.