Non-Durable Consumption and Housing Net Worth in the Great - - PowerPoint PPT Presentation

Non-Durable Consumption and Housing Net Worth in the Great Recession: Evidence from Easily Accessible Data

Greg Kaplan

University of Chicago Kurt Mitman IIES Gianluca Violante New York University New Perspectives on Consumption

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 1 /23

Background

1974 1979 1984 1989 1994 1999 2004 2009 2014 −0.1 −0.08 −0.06 −0.04 −0.02 0.02 0.04 0.06 0.08 0.1

Year Logs (1997:Q1=0) Real ND Consumption

1974 1979 1984 1989 1994 1999 2004 2009 2014 −0.2 −0.1 0.1 0.2 0.3

Year Logs (1997:Q1=0) Real House Prices Boom Bust Bust Boom

• Boom and bust in house prices
• Boom and bust in non-durable consumption

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 2 /23

Background

• U.S. Great Recession 07-09

◮ Sharp drop in house prices ph (∼30 pct) ◮ Durable C expenditures tanked, as in every recessions ◮ Unusually large drop in non-durable consumption (ND-C)

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 3 /23

Background

• U.S. Great Recession 07-09

◮ Sharp drop in house prices ph (∼30 pct) ◮ Durable C expenditures tanked, as in every recessions ◮ Unusually large drop in non-durable consumption (ND-C)

• Causal link between ph and ND-C? How large is this elasticity?

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 3 /23

Background

• U.S. Great Recession 07-09

◮ Sharp drop in house prices ph (∼30 pct) ◮ Durable C expenditures tanked, as in every recessions ◮ Unusually large drop in non-durable consumption (ND-C)

• Causal link between ph and ND-C? How large is this elasticity?
• Answer relevant for:

◮ Consumption insurance ◮ Sources of aggregate fluctuations ◮ Policies that mitigate welfare costs of business cycles

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 3 /23

Widely cited answer

• Mian, Rao and Sufi (QJE, 2012)

◮ The relationship is causal (IV approach) ◮ The elasticity of non-durable consumption to changes in the housing share of net worth (HNW shock) is 0.36

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 4 /23

Widely cited answer

• Mian, Rao and Sufi (QJE, 2012)

◮ The relationship is causal (IV approach) ◮ The elasticity of non-durable consumption to changes in the housing share of net worth (HNW shock) is 0.36

• Methodology:

◮ Geographical (county-level) variation ◮ Instrument: Saiz (2010) local housing supply elasticities ◮ ph data: CoreLogic & ND-C data: MasterCard

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 4 /23

Widely cited answer

• Mian, Rao and Sufi (QJE, 2012)

◮ The relationship is causal (IV approach) ◮ The elasticity of non-durable consumption to changes in the housing share of net worth (HNW shock) is 0.36

• Methodology:

◮ Geographical (county-level) variation ◮ Instrument: Saiz (2010) local housing supply elasticities ◮ ph data: CoreLogic & ND-C data: MasterCard

• Limitation: proprietary data, not replicable

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 4 /23

Our contributions

• 1. Replicate the MRS analysis with (more) easily accessible data

and confirm MRS estimates

• ph data: Zillow & ND-C data: Kilts-Nielsen Retail Scanner

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 5 /23

Our contributions

• 1. Replicate the MRS analysis with (more) easily accessible data

and confirm MRS estimates

• ph data: Zillow & ND-C data: Kilts-Nielsen Retail Scanner
• 2. Separation of price and quantity effect in expenditures
• 1/5 of drop in expenditures due to lower prices

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 5 /23

Our contributions

• 1. Replicate the MRS analysis with (more) easily accessible data

and confirm MRS estimates

• ph data: Zillow & ND-C data: Kilts-Nielsen Retail Scanner
• 2. Separation of price and quantity effect in expenditures
• 1/5 of drop in expenditures due to lower prices
• 3. Use CEX Diary to infer elasticity for Total ND-C
• Elasticity of Total ND-C to ∆ph is 20 pct lower than that of

Kilts-Nielsen bundle

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 5 /23

DATA

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 6 /23

Expenditure data

• Kilts-Nielsen Retail Scanner Data (KNRS)

◮ Weekly panel dataset of sales for over 30,000 stores affiliated with about 90 participating retail chains across 55 MSA geographically dispersed across the US ◮ Information on both quantity sold and price charged per unit at UPC (barcode) level ◮ Construct an annual store-level panel of sales

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 7 /23

Expenditure data

• Kilts-Nielsen Retail Scanner Data (KNRS)

◮ Weekly panel dataset of sales for over 30,000 stores affiliated with about 90 participating retail chains across 55 MSA geographically dispersed across the US ◮ Information on both quantity sold and price charged per unit at UPC (barcode) level ◮ Construct an annual store-level panel of sales

• Store-switching a problem?

◮ For continuing stores, not an issue ◮ Shoppers switching from exiting to surviving stores: attenuation bias

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 7 /23

Coverage

Category All Stores (2006) Baseline Sample (2006-09) Dry grocery 37% 37% Frozen foods 8% 8% Dairy 8% 8% Deli 2% 2% Packaged meat 3% 3% Fresh produce 3% 2% Non-food grocery 13% 13% Alcohol 5% 5% Health and beauty aids 14% 14% General merchandise 8% 9% Number of stores 31,093 14,756

• Bundle composed mostly of groceries, cosmetic, and drugs

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 8 /23

Correlation with NIPA PCE in ND goods and services

• Exclude gasoline and energy from NIPA PCE

0.95 1 1.05 1.1 1.15 1.2 0.95 1 1.05 1.1 1.15 1.2 NIPA KNRS

• State-level correlation in 06-09 nominal sales growth: 0.54

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 9 /23

Net Worth data

NW i

t = Hi t + F i t − M i t − Di t

• Financial assets F: as in MRS, we allocate financial assets in FoF

proportionally to the interests and dividend income from county-level IRS-Statistics of Income

• Mortgage debt M and other debt D: as in MRS, we use Equifax

data underlying the FRB-NY Consumer Credit Panel

• Housing H: compute number of houses by county from ACS, and

multiply by Zillow Home Value Index for All Homes

• Aggregate all at CBSA level (level of Saiz instrument)

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 10 /23

Map of US CBSA

Metropolitan and Micropolitan Statistical Areas

• f the United States and Puerto Rico

February 2013

County or equivalent State Metro area Micro area

Number of areas United States 381 metro areas 536 micro areas Puerto Rico 7 metro areas 5 micro areas

• 929 Core Based Statistical Areas (Metro SA + Micro SA)

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 11 /23

CoreLogic vs Zillow House Price Indexes

!"#$%&'"$%!"#$()*+(,$-./0( ( (((((((((((((((((( ( (

%12324 %56324 %52324 %6324 2324 6324 52324 56324 12324 7#8%22 798%22 :-'%22 ;<$%25 ="<%25 >"?%21 79@%21 A"B%21 C#D%2E &B/%2E C#$%2F ;9G%2F 7#8%26 798%26 :-'%26 ;<$%2H ="<%2H >"?%2I 79@%2I A"B%2I C#D%2J &B/%2J C#$%2K ;9G%2K 7#8%52 798%52 :-'%52 ;<$%55 ="<%55 >"?%51 79@%51 A"B%51 C#D%5E L-$"M-GNB(:#/N-8#@ ON@@-.(:#/N-8#@

!"#$%&'"$%!"#$()*+(,$-./01(234"(5678(( ( (((((((((((((((((( ( (

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

• CoreLogic: repeat-sale index
• Zillow: hedonic price index, includes new constructions

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 12 /23

METHODOLOGY

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 13 /23

Regression specification

• Housing net-worth shock:

∆HNW i

06−09 = ∆ log pi 06−09 ×

• Hi

06/NW i 06

• MRS statistical model:

◮ First stage: ∆HNW i

06−09 = α0 + α1SaizElasti + ηi 06−09

◮ Second stage: ∆ log Cs,i

06−09 = β0 + β1

• ∆HNW i

06−09 + ǫs,i 06−09

• Weight obs. by store sales in 06 & cluster S.E. at CBSA level

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 14 /23

A visual of the first stage

−.7 −.6 −.5 −.4 −.3 −.2 −.1 .1 Log change in Housing Net Worth, 2006−2009 .5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Saiz Housing Supply Elasticity

• Nonlinear relationship
• We use a quartic in the Saiz-elasticity in the first stage

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 15 /23

RESULTS

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 16 /23

Elasticity of ND expenditures to HNW shock

• Dep. var: ∆ log Cs,i

CBSA 2006-09 OLS IV IV (linear) ∆HNW i 0.239** 0.361** 0.405** (0.029) (0.077) (0.089) N 14,756 12,701 12,701 Clusters 281 181 181 R2 0.024 0.017 0.012

• Remarkably similar to MRS estimate of 0.34-0.38 in spite of

different data on ND-C and house prices

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 17 /23

A structural interpretation of this elasticity β1

• Follow Berger-Guerrieri-Lorenzoni-Vavra (2015)
• Life-cycle model with: β(1 + r) = 1, Cobb-Douglas u(c, h),

deterministic income path, no borr. constraints, no trans. costs

• Elasticity of C to a permanent unexpected change in ph:

∆Cit/Cit ∆ph

t /ph t

= Hit T

τ=t

• 1

1+r

τ−t yiτ + Hit + Ait

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 18 /23

A structural interpretation of this elasticity β1

• Follow Berger-Guerrieri-Lorenzoni-Vavra (2015)
• Life-cycle model with: β(1 + r) = 1, Cobb-Douglas u(c, h),

deterministic income path, no borr. constraints, no trans. costs

• Elasticity of C to a permanent unexpected change in ph:

∆Cit/Cit ∆ph

t /ph t

= Hit T

τ=t

• 1

1+r

τ−t yiτ + Hit + Ait ∆ log Cit =    Hit + Ait T

τ=t

• 1

1+r

τ−t yiτ + Hit + Ait   

• β1≃0.25

× ∆ log ph

t

• Hit

Hi,t + Ait

• Similar to income pass-through coeff. of Blundell et al. (2008)

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 18 /23

Consumption vs Expenditures

• Dep. var: ∆ log Cs,i

real

CBSA, 2006-09 OLS IV ∆HNW i 0.196** 0.298** (0.026) (0.085) N 14,756 12,701 Clusters 281 181 R2 0.016 0.012

• Nominal expenditures deflated through Paasche index
• 20 pct of drop in nominal exp. due to lower prices (0.30 vs 0.36)

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 19 /23

From Kilts-Nielsen bundle to Total ND-C

• Wish to translate estimated elasticity in terms of Total ND-C
• Use CE Diary Survey (Attanasio-Battistin-Ichimura, 05) where

items in KN bundle are better measured log cND

it

= Dt + β′

0Xit + β1 log cKN it

+ εit where X: equivalence scale, family type, age, edu, race, region

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 20 /23

From Kilts-Nielsen bundle to Total ND-C

• Wish to translate estimated elasticity in terms of Total ND-C
• Use CE Diary Survey (Attanasio-Battistin-Ichimura, 05) where

items in KN bundle are better measured log cND

it

= Dt + β′

0Xit + β1 log cKN it

+ εit where X: equivalence scale, family type, age, edu, race, region

• NIPA ND goods (excl. energy): KN goods + clothing and footware,

tobacco, books, newspaper and magazines

• NIPA ND goods & services: ND goods + food away from home,

clothing services, entertainment, communication, and transportation services

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 20 /23

From Kilts-Nielsen bundle to Total ND-C

• Wish to translate estimated elasticity in terms of Total ND-C
• Use CE Diary Survey (Attanasio-Battistin-Ichimura, 05) where

items in KN bundle are better measured log cND

it

= Dt + β′

0Xit + β1 log cKN it

+ εit where X: equivalence scale, family type, age, edu, race, region

• NIPA ND goods (excl. energy): KN goods + clothing and footware,

tobacco, books, newspaper and magazines

• NIPA ND goods & services: ND goods + food away from home,

clothing services, entertainment, communication, and transportation services

• Result: elasticity of total ND to KN bundle is 0.7 − 0.9

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 20 /23

Elasticity of ND expenditures to housing wealth

• Dep. var: ∆ log Cs,i

CBSA 2006-09 OLS IV OLS IV ∆ log Hi 0.124** 0.183** (0.019) (0.038) ∆ log

• Hi − Mi

0.072** 0.121** (0.011) (0.025) N 14,756 12,701 13,724 11,745 Clusters 281 181 229 171 R2 0.021 0.017 0.021 0.012

• More intuitive elasticities
• Elast. wrt to housing equity smaller b/c changes are larger

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 21 /23

Taking stock

• Replication of MRS based on alternative and more accessible

data largely confirmed their empirical findings

• Useful ‘moment’ to match to validate structural models

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 22 /23

Taking stock

• Replication of MRS based on alternative and more accessible

data largely confirmed their empirical findings

• Useful ‘moment’ to match to validate structural models
• Our preferred way to state the key quantitative finding:
• 1. The elasticity of total real ND exp. to housing equity is 0.08
• 2. Given an aggregate drop in housing equity of 50 pct, the

implied drop in aggregate ND-C is 4% (half of the total)

• 3. Corresponding annual MPC for ND-C out of housing equity is:

∆Ct Ct = 0.08 × ∆Heq

t

Heq

t

→ MPCHeq = 0.08 × Ct Heq

t

• ∽0.37

= 0.03

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 22 /23

THANK YOU!

Kaplan-Mitman-Violante, ”Consumption and Housing”

• p. 23 /23