Discussion of [2mm] Agarwal, Chua, Ghosh and Song ...
Transcript of Discussion of [2mm] Agarwal, Chua, Ghosh and Song ...
Discussion of
Agarwal, Chua, Ghosh and Song:Consumption and Portfolio Rebalancing Response
of Households to Monetary Policy
Jirka Slacalek
European Central Bankwww.slacalek.com
Monetary Economics and Reality, Helsinki, October 2019
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Key contributions
I Estimate how households rebalance portfolios after cut in interest rates
I Focussing on monetary policy passthrough for term depositors (in India)
I Results
For 100 rupees withdrawn from bank deposits:
I 41 rupees invested in risky assetsI 3 rupees spent on consumptionI Rest: leakage (not covered in the data)I Depositors without auto-renewal accounts and with low liquid savings more
responsive
Flourishing work on MP transmission channels in HANK
Monetary transmission to individual consumption
Direct effects (PE) Indirect effects (GE)
Intertemporal Substitution Income Effects Labor IncomeAsset Prices/Returns
RANK model
• Woodford • Gali • Gertler
Capital Gains Dividends/ProfitsStandard Income Effects through Interest Rates
Income Effects through Mortgage Rates
• Eichenbaum, Rebelo, Wong • Berger, Milbradt, Tourre, Vavra • Wong • Beraja, Fuster, Hurst, Vavra • McKay, Wieland • Greenwald • Hedlund, Karahan, Mitman,
Ozkan • Garriga, Kydland, Sustek
Valuation Effects from Inflation (Fisher Effects)
Fiscal Policy
• Auclert • Sterk,Tenreyro • Doepke,
Schneider • Gottlieb
• Gornemann, Kuester, Nakajima
• Alves, Kaplan, Moll, Violante
• Kekre, Lenel • Werning
• Kaplan, Moll, Violante
• McKay, Nakamura, Steinsson
• Auclert, Rognlie, Straub
• McKay, Reis • Hagedorn,
Manovskii, Mitman
Level
• Kaplan, Moll, Violante
• Luetticke • Auclert • Auclert, Rognlie,
Straub • Werning • Farhi, Werning • Hagedorn, Luo,
Manovskii, Mitman • Bilbiie • TANK model
Risk
• Gornemann, Kuester, Nakajima
• Acharya, Dogra
• Holm • Ravn, Sterk • Werning
• Auclert • Kaplan,
Moll, Violante
• Kaplan, Moll, Violante
• Broer, Hansen, Krusell, Oberg
• Werning • Bilbiie
Moll (2019)
India, May–Oct 2016: 75-bp fall in interest rate“. . . supporting our hypothesis of investors reaching for yield in a period of low interest rate”
But interest rate declines from 7.75% to 7%
35
Figure 1. India’s Interest Rate
Note: This figure shows the bank rate (discount rate) determined by India’s central bank (Reserve Bank
of India), as well as the 1 year Term Deposit rate (with no early withdrawals allowed). The former is
the rate in which the Reserve Bank of India lends money to private bank, while the latter is the rate in
which the private bank pays its term depositors.
India, May–Oct 2016: 75-bp fall in interest rate“. . . supporting our hypothesis of investors reaching for yield in a period of low interest rate”
But interest rate declines from 7.75% to 7%
Identification
I Exploit within-borrower variation. . .
I . . . of term deposits around IR adjustment (after large monetary policy cut)
I Similar to Di Maggio et al. (2017)
I IR change at expiry of term deposit unexpected when they start deposit
Estimation method: Diff in diff
I Compare behavior of depositors with early vs late maturity
I eg, for May 2016:
I Treatment group: Depositors with expired term deposits in MayI Control group: Depositors with expired term deposits in June, July, August
Estimation results
I Interest rate elasticity of C = −1 [Lower end of typical estimates]⇒ Modest direct effects of MP
I Elasticity of rebalancing into risky assets = −6.5⇒ Larger indirect effects of MP
I Stronger reaction of depositors without auto-renewals
I Interest elasticity declines from −3 to −0.5 across deciles of liquid savings
Estimation results
I Interest rate elasticity of C = −1 [Lower end of typical estimates]⇒ Modest direct effects of MP
I Elasticity of rebalancing into risky assets = −6.5⇒ Larger indirect effects of MP
I Stronger reaction of depositors without auto-renewals
I Interest elasticity declines from −3 to −0.5 across deciles of liquid savings
Estimation results
I Interest rate elasticity of C = −1 [Lower end of typical estimates]⇒ Modest direct effects of MP
I Elasticity of rebalancing into risky assets = −6.5⇒ Larger indirect effects of MP
I Stronger reaction of depositors without auto-renewals
I Interest elasticity declines from −3 to −0.5 across deciles of liquid savings
Estimation results
I Interest rate elasticity of C = −1 [Lower end of typical estimates]⇒ Modest direct effects of MP
I Elasticity of rebalancing into risky assets = −6.5⇒ Larger indirect effects of MP
I Stronger reaction of depositors without auto-renewals
I Interest elasticity declines from −3 to −0.5 across deciles of liquid savings
Relation to other work
I Risk-taking channel of monetary policy
I Existing work on banks, mutual funds
I Little evidence on household sector
I Estimation similar to Di Maggio et al. (2017): Effects of ARM mortgagepayments on car purchases
I Artavanis, Paravisini et al. (2019): Effects of (large) uncertainty shocks on earlydeposit withdrawals (Greece), daily data
Relation to other work
I Risk-taking channel of monetary policy
I Existing work on banks, mutual funds
I Little evidence on household sector
I Estimation similar to Di Maggio et al. (2017): Effects of ARM mortgagepayments on car purchases
I Artavanis, Paravisini et al. (2019): Effects of (large) uncertainty shocks on earlydeposit withdrawals (Greece), daily data
Relation to other work
I Risk-taking channel of monetary policy
I Existing work on banks, mutual funds
I Little evidence on household sector
I Estimation similar to Di Maggio et al. (2017): Effects of ARM mortgagepayments on car purchases
I Artavanis, Paravisini et al. (2019): Effects of (large) uncertainty shocks on earlydeposit withdrawals (Greece), daily data
Relation to other work
I Risk-taking channel of monetary policy
I Existing work on banks, mutual funds
I Little evidence on household sector
I Estimation similar to Di Maggio et al. (2017): Effects of ARM mortgagepayments on car purchases
I Artavanis, Paravisini et al. (2019): Effects of (large) uncertainty shocks on earlydeposit withdrawals (Greece), daily data
Relation to other work
I Risk-taking channel of monetary policy
I Existing work on banks, mutual funds
I Little evidence on household sector
I Estimation similar to Di Maggio et al. (2017): Effects of ARM mortgagepayments on car purchases
I Artavanis, Paravisini et al. (2019): Effects of (large) uncertainty shocks on earlydeposit withdrawals (Greece), daily data
Comment 1: How representative of population?
I Sample: Clients of top-four Indian bank
I Nine (large) cities
I How does your sample compare to Indian population by income, wealth, . . . ?
I Is it skewed toward rich, older, urban individuals? More likely to rebalance?
Comment 2: What is the aggregate size of this effect?
I 80 percent of households in India have bank accounts
I What fraction has term deposits?
I How much of total assets do term deposits make up?
I What fraction of accounts expires every quarter?
Household portfolios, India: Low share of (total) financial assets
12 CHAPTER 1. AN INTERNATIONAL COMPARATIVE ANALYSIS
Figure 1.1Country comparison: Allocation of household assets and liabilities
As documented in Badarinza, Balasubramaniam and Ramadorai (2016b), we compare the average alloca-
tions of household assets across countries. The data sources are the All India Investment and Debt Survey
(2012 wave), the Chinese Household Finance Survey (CHFS, 2012 wave), the Townsend Thai Survey (TTS,
2012), the US Survey of Consumer Finances (SCF, 2010 wave), the UK Wealth and Assets Survey (WAS, 2012
wave), the Australian Household, Income and Labour Dynamics Survey (HILDA, 2010 wave), and the Eu-
rosystem Household Finance and Consumption Survey (HFCS). Real estate includes land, buildings and
other constructions owned by the household, for residential, commercial or vacation purposes. Durable
goods include vehicles and other transport equipment, valuables, works of art and carpets, livestock and
poultry, agricultural machinery, and other farm and non-farm equipment. Gold includes jewellery, bullion,
ornaments and coins. Financial assets include bank deposits, publicly traded shares and government se-
curities, mutual funds, managed accounts, and loans receivable by the household. Retirement accounts
include private pension accounts, provident funds, annuity certificates, and life insurance accounts. Mort-
gage debt includes loans using land or real estate as collateral. Other secured debt includes loans secured
by a third party, and loans using crops, shares of companies, government securities, or insurance policies
as collateral. Gold loans use bullion and ornaments as collateral. Unsecured debt includes all loans classi-
fied as personal security, which are not backed up by any collateral, such as unsecured loans from money
lenders, loans from family and friends, credit cards, and overdraft facilities. We compute weighted pop-
ulation averages across households using representative population weights, as indicated in each survey.
0%
20%
40%
60%
80%
100%Assets
India
China
Thaila
ndUSA UK
Austra
lia
Germ
any
Real estateDurable goodsGoldFinancial assetsRetirement accounts
0%
20%
40%
60%
80%
100%Liabilities
India
China
Thaila
ndUSA UK
Austra
lia
Germ
any
Mortgage debtGold loansOther secured debtUnsecured debt
Ramadorai et al., 2017: “Indian Household Finance” report
Household portfolios, India: Low share of fin assets over life cycle
F I G U R E 4 . Allocation of Household Assets and Liabilities Along the Life Cycle
This figure reports the average shares of different assets and debt classes, relative to total household wealth and total household liabilities, respectively. The data sources are the AIDIS (2012 wave), the Eurosystem’s HFCS (2010 wave), the US SCF (2010 wave), the CHFS (2012 wave), the Australian HILDA (2010 wave), and the UK WAS (2012 wave). For consistency, in the CHFS, we classify business loans originated with banks as secured loans and informal private loans as unsecured. We compute averages across households using population weights, as indicated in each survey.
Panel A The assets side of the household balance sheet
0%< 35 35–44 45–54 55–65 > 65
20%
40%
60%
80%
100%Fin. assets and pensions
Perc
ent
Perc
ent
Perc
ent
Age (Years)
IndiaChinaUSA
UKAustraliaGermany
0%< 35 35–44 45–54 55–65 > 65
20%
40%
60%
80%
100%Durable goods and gold
Age (Years)
IndiaChinaUSA
UKAustraliaGermany
0%< 35 35–44 45–54 55–65 > 65
20%
40%
60%
80%
100%Real estate
Age (Years)
IndiaChinaUSA
UKAustraliaGermany
(Figure 4 Continued)Badarinza, Balasubramanian, Ra-madorai, 2017: “Indian FinancialLandscape”
I Financial assets make up 5% oftotal
I Term deposits (small?) fractionof financial assets
I Can still be important forsubpopulations
I Or as fraction of consumption
Figure taken from: Badarinza, Balasubra-
manian, Ramadorai, 2017: “Indian Financial
Landscape”
Comment 3: Do more on heterogeneity across households
1. Which households hold automatic renewal accounts?
2. Size of shocks
I Does eg deposit–income ratio matter?I Do 25 bp policy changes matter less than 75 bp? [Inattention, . . . ]
3. Role of adjustment costs
I How do they vary across households?
I Term deposits: Low adjustment costs (compared to other assets)Which households do / do not adjust?
I Pervasive evidence on inertia in household portfolios, e.g., Ameriks and Zeldes(2004), Brunnermeier and Nagel (2008), Andersen et al. (2018) and others.
Comment 4: Measurement of budget constraint items
I Incomplete measurement of budget constraint: Accounts of individuals (not all Hhmembers) with one bank
I Cover subset of budget constraint items:ATM withdrawals, credit/debit card spending, investments in mutual funds andequities
I 44 % of withdrawals are accounted for, 56 % leakage
I 5 % Leakage for auto-renewals, huge leakage for rest
I Can you do more on to what extent this incomplete measurement matters forestimates?