Do households care about cash? Exploring the heterogeneous effects of India's demonetization

Abstract

The recent demonetization exercise in India is a unique monetary experiment that made 86 percent of the total currency in circulation invalid. In a country where the currency in circulation constitutes 12 percent of GDP, the policy turned out to be a purely exogenous macroeconomic shock that affected all agents of the economy. This paper documents the impact of this macroeconomic shock on households. By construction, the policy helped households with bank accounts in disposing of the demonetized cash. We use a new household-level data set to tease out the effects of this policy on households with no bank accounts relative to households with bank accounts. Our results show that the impact of demonetization on household income and expenditure has been transient. We find that households with no bank accounts experienced a significant decrease in expenditure in November and December-2016. We find a slight decline in income in November-2016 but the effect dissipates thereafter. There is significant heterogeneity in the impact across households in different asset quartiles. We also show evidence of recovery of household finances whereby households were able to smooth out consumption during the post-demonetization period. However, this recovery phase is associated with an increase in household borrowing from different sources. In particular, informal borrowing (e.g. money lenders and shops) increased during this period.

Introduction

On November 8, 2016 the Prime Minister of India, Mr. Narendra Modi, announced that higher denomination notes (500 and 1000 rupee notes) would cease to be legal tender from the midnight of that day. The demonetized notes comprised 86 percent of the total currency in circulation. The demonetization can be characterized as an exogenous macroeconomic shock impacting macro fundamentals, but its effects were expected to percolate to the micro level since cash plays an essential role in the day to day transactions of Indian households and businesses.¹ Moreover, the informal sector in India is large contributing 43.2 percent to Gross Value Added and employing more than 80 percent of the labor force.² Given the large size of the informal sector, a cash shortage due to demonetization is expected to impact the majority of the players in the Indian economy. In this paper, we focus on one such important player in the economy: the households. We look at household level data to tease out the effects of this sudden liquidity shock on income and expenditure. We then delve deeper to ask if demonetization had a heterogeneous effect on households. Although, the costs of demonetization appear almost immediately while the benefits are expected to be seen in a more medium term horizon with the broadening of the tax base and more digitization of payments, in this paper, we mainly focus on the short term costs of demonetization given lack of longer term time series for analyzing the benefits. However, we provide some evidence of the recovery phase that followed after demonetization. In what follows, we quantify the impact of the policy on households, and also uncover interesting and important dimensions of heterogeneity in the data. Before we provide more details of the policy in the next section, we present a brief review of the literature to place our study in the larger context of liquidity constraints.

Our paper relates to the literature on household liquidity, income shocks, and consumption smoothing. Several empirical studies on household saving and consumption have examined the importance of liquidity constraints ([Zeldes, 1989], [Jappelli, 1990], [Runkle, 1991], [Jappelli et al., 1998]) and quite a few of these papers used data on credit card usage and changes in borrowing limits. Zeldes (1989) partitioned households in his sample according to financial wealth relative to income and total wealth relative to income and defined a constrained household as one with assets worth less than 2 months of income. Souleles (1999) identified credit constrained households by holdings of liquid wealth relative to earnings. This classification of households allowed him to document that the consumption of non-durable goods for credit constrained families (the bottom 15% of the liquid wealth-to-earnings distribution) is sensitive to predictable changes in earnings, whereas non-durable consumption for unconstrained households (the top 25th percentile of the liquid wealth-to-earnings distribution) is not sensitive to these anticipated changes. Gross and Souleles (2002) examined households’ responses to exogenous changes in the borrowing limit on credit cards. They found that, on average, consumers increase their debt holdings by 10% to 14% of the increase in borrowing. Investigating the effects of the US Supreme Court decision that deregulated bank credit card interest rates in December 1978, Zinman (2003) compared consumers’ acquisition and usage of credit cards between states that mandated binding usury limits before the court decision and states that were unaffected by deregulation. His results suggest that households who seem to be credit constrained used the easier access to credit to acquire credit cards and borrow frequently on their new credit cards. More recently Krueger and Perri (2011) showed that the effects of labor income shocks on consumption are modestly persistent. This is so because consumption can be well insured using simple unsecured borrowing and savings. In the context of demonetization in particular, Chodorow-Reich et al. (2020) presented a model in which households hold cash for both cash-in-advance (CIA) needs and tax evasion purposes. The model predicts that in the presence of downward wage rigidity, a forced conversion of cash into less liquid assets due to demonetization results in a decline in economic activity.

Our definition of liquidity constrained is different from the above literature because India is more cash-based than advanced economies. The demonetization exercise reduced the amount of cash that individuals held, and we seek to explore the adjustment process that followed thereafter. Given the nature of the demonetization event, having a bank account was quite essential because it facilitated exchange of the demonetized banknotes. We compare households with bank accounts to those without bank accounts. In addition, we differentiate households according to asset holdings. To this end, we construct an asset index and look at the impact of the liquidity shock by quartile of this index.

There is an emerging strand of literature that specifically studies the demonetization exercise in India. Aggarwal and Narayanan (2017) studied the impact on the agricultural sector while Dash et al. (2017) showed how demonetization has led households to save through more formal channels. RBI (2017) analyzed broad macroeconomic trends in the aftermath of demonetization and Behera et al. (2017) studied the impact on the financial sector. In a more recent paper Chodorow-Reich et al. (2020) used geographic distribution of demonetized notes to draw causal inferences on economic activity. They showed that districts facing more severe demonetization had larger relative reductions in economic activity, faster adoption of payment alternatives, and lower bank credit growth. They concluded that on a cumulative basis, economic activity (night lights based output and employment) contracted by 2 percentage points in 2016Q4 (calendar quarter) relative to their counterfactual growth paths. However, these effects dissipated over the next few months. Kurosaki (2019) used a panel data set on registered and unregistered manufacturing firms to show that even after the demonetization shock, both types of firms still remained cash-dependent. Contrary to the other research on demonetization, Chanda and Cook (2019) using night light data showed that districts with higher deposit growth during demonetization experienced higher levels of economic activity. They used the Consumer Pyramids data to look at the effect on household income and expenditure with respect to deposit growth across districts and found similar results.

Our study, in contrast uses a cleaner identification strategy based on the ownership of bank accounts (which is orthogonal to the demonetization shock) to estimate the effects of demonetization on household balance sheet items. We also analyze heterogeneous consequences for households which is not done in the other papers. We confirm the results obtained in previous studies that the aggregate macro impact was transient. In contrast to other studies, however, we document the impact of the macro-level demonetization shock at the micro level. Our results show that the impact of demonetization on household income and expenditure has been transient. We find that households with no bank accounts experienced a significant decrease in expenditures in November and December 2016. We find a slight decline in income in November-2016 but the effect dissipates thereafter. These effects however differ for households across different asset quartiles. We also show evidence of recovery of household finances whereby households were able to smooth out consumption during the post-demonetization period. This recovery phase is associated with an increase in household borrowing from different sources, primarily for the purpose of consumption. In particular, informal borrowing (money lenders, retail shops) increased substantially during this period.

The rest of the paper is organized as follows. Section 2 presents background details on demonetization. Section 3 presents the data and descriptive statistics, and also discusses due-diligence we have done on the dataset used. Section 4 discusses the empirical strategy. Section 5 presents the results, and Section 6 concludes.