How Does Monetary Policy Affect Income and Wealth ... · The views expressed in this presentation are those of the authors and do not necessarily re ect the views of the European

How Does Monetary Policy Affect Income andWealth Inequality? Evidence from Quantitative

Easing in the Euro Area

Michele Lenza Jirka Slacalek

European Central Bank

Challenges in Understanding the Monetary Transmission Mechanism

Warsaw, 22 March 2019

The views expressed in this presentation are those of the authors and do not necessarily

reflect the views of the European Central Bank and the Eurosystem.

Lenza and Slacalek (ECB) EABCN Conference 22 March 2019 1 / 20

Motivation and main results

Recent (academic and public) debate on impact ofquantitative easing on inequality

Widely diverging perspectives on how QE may affect inequality (see alsoColciago, Samarina and de Haan, 2018):

I QE boosted asset prices and financial wealth, it “made the rich richer’’(eg FT, Oct 21, 2014)

I However, QE also boosted house prices: these gains are more widely spread,as homeowners more evenly distributed than stock-holders

I Expansionary mon policy reduces unempl, benefits poorer households most

ECB has since 2015 undertaken quantitative easing (QE)(“Asset Purchase Programmes” - APP)



Recent (academic and public) debate on impact ofquantitative easing on inequality

Widely diverging perspectives on how QE may affect inequality (see alsoColciago, Samarina and de Haan, 2018):

I QE boosted asset prices and financial wealth, it “made the rich richer’’(eg FT, Oct 21, 2014)

I However, QE also boosted house prices: these gains are more widely spread,as homeowners more evenly distributed than stock-holders

I Expansionary mon policy reduces unempl, benefits poorer households most

ECB has since 2015 undertaken quantitative easing (QE)(“Asset Purchase Programmes” - APP)



Question and method

What are the effects of QE on inequality in the euro area?

Step 1: Aggregate data

Multi-country VAR with, among other things, aggregate unemployment,wages and asset prices ⇒ Impulse responses to QE shock

Step 2: Household-level data, Household Finance and ConsumptionSurvey

Transpose IRFs over household-level data ⇒ Estimate effects of QE on

wealth and income inequality (Gini index)

Income (composition and earning heterogeneity channel); Wealth(composition channel)



Question and method










Question and method










Question and method










Existing literature

VARs with income / consumption Ginis:Coibion et al. (JME, 2017); Mumtaz and Theophilopoulou (EER, 2017)

I No wealth inequality, don’t estimate effects of nonstandard MP

Household wealth portfolios, inflation and asset prices:Doepke and Schneider (JPE, 2006); Adam and Zhu (JEEA, 2016); Adam and Tzamourani (EER, 2016); Doepke et al.

(2016)

I Assume hypothetical scenarios, eg “10% increase in price level”

Model-based simulations:Casiraghi et al. (2018) [BdI]; Bunn et al. (2018) [BoE]

I More calibrated than estimated

⇒ Little quantitative, estimated work on effects of QE on inequality



Existing literature

VARs with income / consumption Ginis:Coibion et al. (JME, 2017); Mumtaz and Theophilopoulou (EER, 2017)

I No wealth inequality, don’t estimate effects of nonstandard MP

Household wealth portfolios, inflation and asset prices:Doepke and Schneider (JPE, 2006); Adam and Zhu (JEEA, 2016); Adam and Tzamourani (EER, 2016); Doepke et al.

(2016)

I Assume hypothetical scenarios, eg “10% increase in price level”

Model-based simulations:Casiraghi et al. (2018) [BdI]; Bunn et al. (2018) [BoE]

I More calibrated than estimated

⇒ Little quantitative, estimated work on effects of QE on inequality



Main Results

One year after the occurrence of the QE shock:

QE reduces income inequality⇒ Key role of the earnings heterogeneity channel (extensive margin,transition out of unemployment)

Wealth inequality is largely unchanged (in background slidestoday)


The aggregate effects of QE: multi-country VAR

Outline

1 The aggregate effects of QE: multi-country VAR

2 Distributing the QE effects to individual households: HFCS

3 Robustness checks

4 Conclusions



Step 1: Multi-country VAR to estimate aggr effects of QE

y t = C + B1yt−1 + · · ·+ Bpyt−p + εt

εt = N(0,Σ)

Mix of EA and country-level variables; 4 countries: DE, FR, IT, ES⇒ Common MP + country heterogeneity in responses

Variables yt (quarterly, 1999Q1–2016Q4, p = 5 lags)I Country-specific: real GDP, GDP defl, wages, unempl, house pricesI EA: short- and long-term interest rates, stock pricesI US: GDP, short-term interest rates

Large dimension ⇒ Bayesian estimation (Litterman, 1979; Doan et al., 1984; Sims,

1992; Banbura et al., 2010; Giannone et al., 2015) Uses of multi-country BVARs for the euro area




y t = C + B1yt−1 + · · ·+ Bpyt−p + εt

εt = N(0,Σ)








y t = C + B1yt−1 + · · ·+ Bpyt−p + εt

εt = N(0,Σ)







QE ”scenario”

1 Identify exogenous asset purchase shock with zero and signrestrictions (Baumeister and Benati, 2013; Arias et al., 2017)

Expansionary QE (APP) shock on impact:I Leaves short-term interest rate unchangedI Decreases long-term interest rates ⇒ Decreases term spreadI Increases real GDP

2 Offset response of EA policy rate via series of standard MP shocks

I Standard MP did not react to offset effects of asset purchases (policyrate remained at lower bound)

I Standard MP shock identified via standard zero (Choleski) restrictions



QE ”scenario”

1 Identify exogenous asset purchase shock with zero and signrestrictions (Baumeister and Benati, 2013; Arias et al., 2017)

Expansionary QE (APP) shock on impact:I Leaves short-term interest rate unchangedI Decreases long-term interest rates ⇒ Decreases term spreadI Increases real GDP

2 Offset response of EA policy rate via series of standard MP shocks

I Standard MP did not react to offset effects of asset purchases (policyrate remained at lower bound)

I Standard MP shock identified via standard zero (Choleski) restrictions



Impulse responses—QE shock

Size of QE shock to term spread scaled to 30 bp on impactIn line with Altavilla et al. (2015) and Andrade et al. (2016)

-6-4

-20

24

Per

cent

(Sto

ck P

rices

)

-.3-.2

-.10

.1P

erce

ntag

e P

oint

s (In

tere

st R

ates

)

0 4 8 12

Quarter after Shock

Short-term Rate Long-term Rate Stock Prices (RHS)

Impulse Responses of Financial Variables (Euro Area)



Impulse responses of some key aggregate variables

UR, HP responses stronger in ES, milder in DE

Link to mortgage / labor market institutions?

-1-.8

-.6-.4

-.2P

erce

ntag

e P

oint

s

0 4 8 12

Quarter after Shock

Germany Spain France Italy

Impulse Response of Unemployment

-.50

.51

1.5

2P

erce

nt0 4 8 12

Quarter after Shock


Impulse Response of House Prices

All other responses roughly as expected (very mild response of pricesand wages)


Distributing the QE effects to individual households: HFCS

Outline



3 Robustness checks

4 Conclusions



Bringing IRFs to HFCS micro data—Income

1 Earnings heterogeneity channel (extensive margin)Distribute aggregate decline in unemployment across people using asimple probit simulation, some unemployed become employed

2 Income composition channel (intensive margin)Income of employed increases in line with the IRF for wages

0

10

20

30

40

50

60

70

80

90

100

1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintile

Source: HFCS 2nd wave. Countries: Euro area countries.

percentages over total income

Income composition

employee income income from self-employment pensions financial income

rental income unemployment benefits transfers



Unemployment simulation—Extensive margin [Ampudia et al. (2016)]

1. Who becomes employed? Probit model

Country (c)-specific at individual level (not Hh):

Pr(Y = 1|X = x) = Φ(x ′c,i βc)

Y empl status, X demographics (gender, edctn, age, mar status, chldrn)

Collect fitted probability to be employed: Yc,i

Simulation: those with an higher Yc,i are more likely to become employed∑newly employed people = aggregate decline in unempl implied by VAR

2. Which wages do newly employed get? Estimate unobserved wages

Income of the newly employed increases as implied by wage regression(Heckman): wage instead of (lower) unempl benefits

The wage is based on demographic characteristics

Technicalities



Unemployment simulation—Extensive margin [Ampudia et al. (2016)]



Pr(Y = 1|X = x) = Φ(x ′c,i βc)


Collect fitted probability to be employed: Yc,i

Simulation: those with an higher Yc,i are more likely to become employed∑newly employed people = aggregate decline in unempl implied by VAR


Income of the newly employed increases as implied by wage regression(Heckman): wage instead of (lower) unempl benefits

The wage is based on demographic characteristics

Technicalities



UnemploymentDisproportionate decrease for low income Unemployment rates by income group

-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

GermanyResponse of Unemployment by Income Quintile

-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

SpainResponse of Unemployment by Income Quintile

-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

FranceResponse of Unemployment by Income Quintile

-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

ItalyResponse of Unemployment by Income Quintile



Income inequalityUnempl benefits more generous in DE, FR than in ES and IT

01

23

45

6P

erce

nt

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

GermanyResponse of Income by Income Quintile

01

23

45

6P

erce

nt

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

SpainResponse of Income by Income Quintile

01

23

45

6P

erce

nt

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

FranceResponse of Income by Income Quintile

01

23

45

6P

erce

nt

0 4 8 12

Quarter

Bottom 20% 20-40% 40-60%60-80% Top 20%

ItalyResponse of Income by Income Quintile



EA Income inequalityLower inequality: Gini goes down from 43.07 to 42.86Key importance of extensive margin (Unemp → Emp)

01

23

Per

cent

1

(EUR 9,400)

2

(EUR 19,700)

3

(EUR 29,900)

4

(EUR 44,700)

5

(EUR 95,300)

Growth of Mean Income by Income QuintileExtensive margin(Unemp → Emp)

Intensive margin(wage growth)

Response of mean income 4 quarters after QE shock. Numbers in brackets: Initial levels of mean gross Hh income.


Robustness checks

Outline



3 Robustness checks

4 Conclusions


Robustness checks

Robustness checks on income inequality

Main question: what about the dynamics of financial income?

Issue: lack of good data on financial incomeI Proxy 1: profits for the euro areaI Proxy 2: net property income for the four countries

Local linear projections (Jorda, 2005):How do these variables respond to QE shock?

I ProfitsI Net property income

Using the two proxies of financial income, Gini drops from 43.07 to42.89 (as opposed to 42.86 in the baseline). The result on incomeinequality is unchanged.


Robustness checks








Robustness checks








Conclusions

Outline



3 Robustness checks

4 Conclusions


Conclusions

Conclusions

QE reduces income inequalityI Substantial impact on employment at bottom tail

The effect of QE on wealth inequality is likely to be small


Background Slides

Background slides


Background Slides

Impulse responses of aggregate variables

-6-4

-20

24

Per

cent

(Sto

ck P

rices

)

-.3-.2

-.10

.1P

erce

ntag

e P

oint

s (In

tere

st R

ates

)

0 4 8 12

Quarter after Shock

Short-term Rate Long-term Rate Stock Prices (RHS)

Impulse Responses of Financial Variables (Euro Area)

-.50

.51

1.5

2P

erce

nt

0 4 8 12

Quarter after Shock


Impulse Response of House Prices

-1-.8

-.6-.4

-.2P

erce

ntag

e P

oint

s

0 4 8 12

Quarter after Shock


Impulse Response of Unemployment

-.20

.2.4

Per

cent

0 4 8 12

Quarter after Shock


Impulse Response of Wages


Background Slides

Impulse responses 4 quarters after shock

Substantial heterogeneity across countries

DE FR IT ESpp. change over one year

0

0.5

1

1.5

2House prices

DE FR IT ESchange over one year

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0Unemployment rate

DE FR IT ESpp. change over one year

-0.2

0

0.2

0.4

0.6Wages

LTN, change impact S. P., pp. change after one year-0.4

-0.2

0

0.2

0.4

0.6

0.8EA financial variables


Background Slides

Modelling response of wealth and incomecomponents to QE Back

Table 1 Modeling of Responses of Wealth and Income Components

Wealth / income component Modeling procedure

Real AssetsHousehold's main residence Multiplied with response of house pricesOther real estate property Multiplied with response of house pricesSelf-employment businesses Multiplied with response of stock prices

Financial AssetsShares, publicly traded Multiplied with response of stock prices (in the baseline; robustness: some trading)Bonds Multiplied with response of bond prices (based on long-term rate)Voluntary pension/whole life insurance No adjustmentDeposits No adjustmentOther �nancial assets No adjustment

DebtTotal liabilities No adjustment

Gross IncomeEmployee income Multiplied with response of wages (compensation per employee)Self-employment income Multiplied with response of wages (compensation per employee)Income from pensions No adjustmentRental income from real estate property No adjustmentIncome from �nancial investments No adjustment (in the baseline; robustness: grows by 5%)Unemployment bene�ts and transfers If becomes employed, replace with wage (otherwise no adjustment)

26


Background Slides

Unemployment simulation—Extensive margin [Ampudia et al.

(2016)] Back



Pr(Y = 1|X = x) = Φ(x ′c,i βc)


Collect fitted values Yc,i ; draw uniformly distributed shock εc,i

If εc,i sufficiently below Yc,i ⇒ unempl individual i becomes employed∑newly employed people = aggregate decline in unempl implied by VAR

Repeat many times for different draws of εc,i , average across sims


Income of the newly employed increases as implied by Heckman:They receive wage instead of (lower) unempl benefitsExclusion restrictions: marital status, children


Background Slides

EA unemploymentDisproportionate decrease for low income

-2-1

.5-1

-.5

0

Per

cent

age

poin

ts

1

(40.4%)

2

(14.8%)

3

(9.2%)

4

(3.8%)

5

(2.4%)

Decline in Unemployment Rate by Income Quintile


Background Slides

UnemploymentES: Unemployed affected in all quintiles b/c distributed more evenlyDE: UR strongly skewed toward lowest income quintile Back

010

2030

4050

Per

cent

DE ES FR IT

Unemployment Rate by Income QuintileBottom 20% 20-40% 40-60% 60-80% Top 20%


Background Slides

Bringing IRFs to HFCS micro data—WealthBack

Estimate effects on household-level net wealth using holdings ofhousing wealth, stocks and bonds (in e) Detail

Housing, stock, bonds account for about 80% of value of wealthAssumes no rebalancing of portfolios

Composition of total assets

0

10

20

30

40

50

60

70

80

90

100

Per

cent

of T

otal

EU

R V

alue

1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintileQuintile of Net Wealth

household main residence other real estate self-employment business shares, publicly traded

bonds voluntary pension/life insur. other financial assets depositsLenza and Slacalek (ECB) EABCN Conference 22 March 2019 20 / 20

Background Slides

Wealth inequalityVery small effect: Gini goes down from 68.09 to 68.07

Important to account for house prices Decomposition

[Assumes: no portfolio rebalancing; in line with literature on inertia in Hh portfolios (Ameriks,

Zeldes, 2004; Bilias et al. (2010)]

0.5

11.

52

2.5

Per

cent

1

(€ 1,100)

2

(€ 25,200)

3

(€ 111,400)

4

(€ 225,900)

5

(€ 512,400)

Growth of Median Net Wealth by Net Wealth Quintile

Response of median net wealth 4 quarters after QE shock. Numbers in brackets: Initial levels of median net wealth.


Background Slides

Decomposition of changes in net wealthKey role of housing Back

0.5

11.

52

2.5

Per

cent

Lowest 30% 30-70% 70-95% Top 5%

by Net Wealth Quantile (Mean)Growth of Net Wealth and Its Components

Net Wealth Housing Stocks and Bonds

Response of mean net wealth and its components 4 quarters after QE shock.


Background Slides

Net wealthCaveat: Some increase in wealth above P90, but transitory (see IRF for stock prices)

Lower percentiles: Role of leverage0

.51

1.5

22.

5P

erce

nt

0 4 8 12

Quarter

20-40% 40-60%60-80% Top 20%

GermanyResponse of Wealth by Wealth Quintile

0.5

11.

52

2.5

Per

cent

0 4 8 12

Quarter

20-40% 40-60%60-80% Top 20%

SpainResponse of Wealth by Wealth Quintile

0.5

11.

52

2.5

Per

cent

0 4 8 12

Quarter

20-40% 40-60%60-80% Top 20%

FranceResponse of Wealth by Wealth Quintile

0.5

11.

52

2.5

Per

cent

0 4 8 12

Quarter

20-40% 40-60%60-80% Top 20%

ItalyResponse of Wealth by Wealth Quintile


Other robustness checks

Robustness

Local linear projections (Jorda, 2005):How do other variables respond to QE shock?

I Holdings of wealth components (flow of funds)I ES local house pricesI ES local house prices: IRF vs levelI Profits / financial income

Uniform employment probability

Same VAR response in all countries

Portfolio rebalancing—some trading in stocks:Buy 15% of your stock holdings



Local linear projection:ES holdings of wealth components (flow of funds) Back

Total fin assets ↑≈ 5–10%; stocks ↑ by a lot (≈ 15%), debt ↓ a bit

0 5 10 15−20

−10

0

10

20

30Total Financial Assets

0 5 10 15−5

0

5

10

15

20Currency and Deposits

0 5 10 15−80

−60

−40

−20

0

20

40

60Debt Securities

0 5 10 15−40

−20

0

20

40

60Stocks and Other Equity

0 5 10 15−10

−5

0

5

10

15

20Insurance, Pension Schemes



Local linear projection: ES regional house pricesBack

Some, but not overwhelming heterogeneity

0 2 4 6 8 10 12 14−5

0

5

10

15

20

25

Quarter

Per

cent



ES regional house prices: IRF vs level Back

Positive relationship b/w level and response of HP

800 1000 1200 1400 1600 1800 2000 2200 2400 26000

2

4

6

8

10

12

Average price per squared meter

Per

cent

age

incr

ease

a y

ear

afte

r th

e sh

ock



Local linear projection: Profits ↑ by 5% Back

1 3 5 7 9 11-6

-4

-2

0

2

4

6

8

10

12



Local linear projection: Net property ↑ by 4% to20% Back

1 3 5 7 9 11-10

-5

0

5

10

15

20Germany

1 3 5 7 9 11-10

-5

0

5

10

15France

1 3 5 7 9 11-50

-40

-30

-20

-10

0

10

20

30

40

50Spain

1 3 5 7 9 11-20

-15

-10

-5

0

5

10

15

20

25Italy



Robustness: Uniform employment probabilityBaseline IRFs (Solid) vs IRFs under uniform probability of getting employed (Dashed) Back

-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5


-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5


-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5


-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5




Robustness: Same VAR response in all countriesBaseline IRFs (Solid) vs IRFs restricted to be the same across countries (Dashed) Back

-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5


-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5


-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5


-2.5

-2-1

.5-1

-.50

Per

cent

age

Poi

nts

0 4 8 12

Quarter

Q1 Q2 Q3 Q4 Q5Q1 Q2 Q3 Q4 Q5




Robustness: Financial income ↑ by 5%Financial income matters most in the upper tail Back

01

23

4

Per

cent

1

(EUR 9,400)

2

(EUR 19,700)

3

(EUR 29,900)

4

(EUR 44,700)

5

(EUR 95,300)

Growth of Mean Income by Income QuintileExtensive margin(Unemp → Emp)

Intensive margin(wage growth)

Financialincome

Numbers in brackets: Initial levels of mean gross Hh income.



Robustness: Holdings of stocks ↑ by 15%Similar overall results Back

High leverage at the bottom

01

23

Per

cent

1

(€ 1,100)

2

(€ 25,200)

3

(€ 111,400)

4

(€ 225,900)

5

(€ 512,400)

Growth of Median Net Wealth by Net Wealth QuintileBaseline Buying Stocks Counterfactual

Numbers in brackets: Initial levels of median net wealth.



Net nominal positions



Net interest rate exposure—Auclert (2017)Net interest rate exposure = maturing assets - maturing liabilities

Maturing assets = 25% of value of mutual funds, bonds, shares, managedaccounts, money owed to households, other assets + 100% of deposits

Maturing liabilities = 100% outstanding balance of adjustable-rate mortgages +100% outstanding balance of other non-collateralized debt



Nonstandard vs Standard MP

Targeting the same peak GDP response, VAR gives:30 bp change in term spread ≈ 100 bp change in policy rate

BUT also qualitative differences (ZLB, differential effects on prices ofspecific assets, . . . )



Uses of Multi-Country BVARs

Altavilla, Giannone and Lenza (IJCB, 2016)Effects of OMT and standard policy in DE, ES, FR and IT

Mandler, Scharnagl and Volz (WP, 2016)Effects of standard policy in DE, ES, FR and IT

Angelini, Lalik, Lenza and Paredes (IJF, 2019)Evaluation of conditional and unconditional forecasts for DE, ES, FRand IT

Capolongo and Pacella (mimeo BdI, 2019)Inflation forecasts for DE, ES, FR and IT

Back


Documents

How Does Monetary Policy Affect Income and Wealth ... · The views expressed in this presentation are those of the authors and do not necessarily re ect the views of the European