The Effect of the Recovery Act on Consumer Spending

Bill Dupor (FRB of St Louis) Marios Karabarbounis (FRB of Richmond) Marianna Kudlyak (FRB of San Francisco) M. Saif Mehkari (University of Richmond)

1 / 28

The analysis and conclusions set forth do not reflect the views of the institutions to which the authors are affiliated.

1 / 28

The Research Question: If the government purchases $1 worth of goods by how many dollars does private consumption increase or decrease?

– Very old question but literature still lacks consensus • Ramey and Shapiro (1998), Blanchard and Perotti (2002), Gali et al. (2007) • consumption multipliers from slightly negative to 0.5.

2 / 28

What we do We analyze the effect of the Recovery Act on consumer spending. Very large program! Spending component around $228 Billion.

Million $ 0

(0-10]

(10-20] (20-50]

(50-100]

(100-500] >500

Notes: Recovery Act Data (2009-2012): $ amounts awarded per county (Authors’ calculations). 3 / 28

What is new? • We use regional variation to estimate the consumption response to

gov. spending. — Large literature using regional variation to study effects of gov. programs (Chodorow-Reich et al. (2012), Conley and Dupor (2013), Nakamura and Steinsson (2014)). — We look explicitly at local consumer spending: combine micro-level data on retail purchases (Nielsen Store/HomeScan) and auto purchases (auto balances from Equifax).

4 / 28

What is new? • We use regional variation to estimate the consumption response to

gov. spending. — Large literature using regional variation to study effects of gov. programs (Chodorow-Reich et al. (2012), Conley and Dupor (2013), Nakamura and Steinsson (2014)). — We look explicitly at local consumer spending: combine micro-level data on retail purchases (Nielsen Store/HomeScan) and auto purchases (auto balances from Equifax).

• Translate estimates using a HA-NK model with many regions — aggregate local to national (Nakamura and Steinsson (2014), Beraja, Hurst and Ospina (2016)). — Novel part: we also allow for heterogeneity/incomplete markets within regions→ heterogeneity in MPC’s.

4 / 28

Our preliminary findings •• The spending component of the Recovery Act increased local consumer spending. 1% increase in gov. spending increased the growth rate of • retail spending by 0.033 p.p. • auto spending by 0.060 p.p

5 / 28

Our preliminary findings •• The spending component of the Recovery Act increased local consumer spending. 1% increase in gov. spending increased the growth rate of • retail spending by 0.033 p.p. • auto spending by 0.060 p.p

•• Our quantitative model implies that • Local variation in wages → positive local consumption multiplier. • Monetary policy response affects both the local and the aggregate consumption multiplier.

5 / 28

Data Sources and Empirical Analysis

5 / 28

A narrative/quantitative IV approach

To overcome endogeneity we find components of the Act allocated in a way that was uncorrelated with local business cycle • do a detailed reading of the Act, federal codes and regulations cited

in the act and implementation guidances written by agencies tasked with allocating funds (100s of pages!) • catalogue 7 programs that are exogenous to local business cycle • Examples: water quality assistance grants and Dept. of Education’s

Special Education Fund

6 / 28

Components of the Recovery Act used in the construction of the instrument

Federal Department/Agency

Total Amount Fraction included Authorized in IV ($Billions) (%)

Environmental Protection Agency General Services Administration Department of Transportation Department of Education Department of Energy Department of Justice Department of Defense All other Agencies

6.7 4.8 39.3 71.6 33.3 3.5 4.3 62.3

87.5 98.3 16.7 15.6 43.5 72.4 87.1 0.0

All Departments/Agencies

228.0

20.2

7 / 28

Nielsen HomeScan Data Panel of approximately 60,000 households/ 40,000 stores for period 2002-2013. Information on amount spent/total sales and area of residence/store location (zip code). Store Type Grocery Discount store Warehouse club Drug store Department store Online Shopping Hardware/Home Improv. Dollar Store Apparel Stores

Spending 32.9% 20.5% 8.5% 4.2% 3.9% 3.0% 2.9% 1.7% 1.6%

Convenience store Electronics store Gas mini mart Pet store Restaurant Office supplies store Quick serve restaurants Liquor store Home furnishings

1.5% 1.1% 1.0% 0.8% 0.7% 0.7% 0.6% 0.6% 0.5%

8 / 28

Equifax 5% random sample of all U.S. consumers with credit records: approx. 10 million individuals. Information on household debt: mortgages, credit cards, student loans, and auto loans. Use change in auto balance as a proxy for auto spending. 180 160 140 120 100 80 60 40

# Loans (2010=100) Passenger Car Registration in U.S. (FRED)

20 0

2000

2002

2004

2006

2008

2010

2012

2014

9 / 28

Econometric Model

∆ log Cj = a + β × log Gj + Xj Φ0 + Dj + εj with

∆ log Cj =

2012 X

{log Cj,t − log Cj,2008 }

t=2009

Xj : county’s population, per-capita change in county income between 2008-2012 Dj : state dummy

10 / 28

Retail and Auto Spending Effects of the Recovery Act

10 / 28

Percentage Change in County-Level Retail Spending -.05 .03 .11 .19 .27 .35

Relationship between Recovery Act spending and retail spending, county-level

5

6 7 8 County-Level Recovery Act Spending (in log)

9

11 / 28

1% increase in government spending increases retail spending growth rate between 0.016-0.037 p.p. Spending Category

Retail Spending (Nielsen, Store-level) OLS

Recovery Act Spending Partial F stat. County Controls/State F.E.

# Counties

IV

OLS

IV

0.016 0.037∗ 0.029∗ 0.033∗ (0.020) (0.022) (0.017) (0.017) — No 365

132.4 No 365

— Yes 365

213.4 Yes 365

Note: the Table reports the regression estimates based on Store-level data. We weight by population and we cluster standard errors at the state level. State

12 / 28

1% increase in government spending increases retail spending growth rate between 0.022-0.032 p.p. Spending Category

Retail Spending (Nielsen, HomeScan) OLS

Recovery Act Spending Partial F stat. County Controls/State F.E.

# Counties

IV

OLS

IV

0.022∗ 0.029∗∗ 0.032∗ 0.023 (0.013) (0.014) (0.017) (0.019) — No 525

166.9 No 525

— Yes 524

220.4 Yes 524

Note: the Table reports the regression estimates based on HomeScan data. We weight by population and we cluster standrard errors at the state level.

13 / 28

Total Non-durable elasticity We translate our Nielsen elasticity into a nondurable elasticity. • Nielsen type purchases from the CEX: food at home, alcohol and

beverage, deteregents, cleaning products and other household products, small appliances, and personal care products • Total non-durables: Nielsen + food away from home, a broader set of housing supply purchases, utility bills and public transportation (on average 4.6 times larger). non-durable Nielsen log Ci,t = a + β × log Ci,t + Xi,t Φ0 + εi,t

β = 0.55: the nondurable elasticity is around half our Nielsen coefficient.

14 / 28

Percentage Change in County-Level Auto Spending 0 .1 .2 .3 .4 .5

Relationship between Recovery Act spending and auto spending, county-level

5

6 7 8 County-Level Recovery Act Spending (in log)

9

15 / 28

1% increase in government spending increases auto spending growth rate between 0.045-0.060 p.p. Spending Category

Auto Spending (Equifax) OLS

Recovery Act Spending Partial F stat. County Controls/State F.E.

# Counties

Placebo Regressions

IV

OLS ∗∗∗

IV

0.015 0.061 0.045 0.060∗∗∗ (0.024) (0.023) (0.014) (0.015) — No 2999

205.9 No 2930

∗∗∗

— Yes 2936

162.8 Yes 2872

Heterogeneity

16 / 28

Model

16 / 28

Description of the Economy • i = {1, .., N} regions (N = 2 for this presentation).

17 / 28

Description of the Economy • i = {1, .., N} regions (N = 2 for this presentation). • Each region is an open Huggett economy: continuum of households

making consumption, working, and saving decisions.

17 / 28

Description of the Economy • i = {1, .., N} regions (N = 2 for this presentation). • Each region is an open Huggett economy: continuum of households

making consumption, working, and saving decisions. • Each region produces a final good using local and foreign

intermediate inputs. – 1st spillover: trade across regions

17 / 28

Description of the Economy • i = {1, .., N} regions (N = 2 for this presentation). • Each region is an open Huggett economy: continuum of households

making consumption, working, and saving decisions. • Each region produces a final good using local and foreign

intermediate inputs. – 1st spillover: trade across regions • Federal government supplies a bond. Buys final goods from all

regions and finances local expenditures using federal taxes. – 2nd spillover: federal fiscal authority

17 / 28

Description of the Economy • i = {1, .., N} regions (N = 2 for this presentation). • Each region is an open Huggett economy: continuum of households

making consumption, working, and saving decisions. • Each region produces a final good using local and foreign

intermediate inputs. – 1st spillover: trade across regions • Federal government supplies a bond. Buys final goods from all

regions and finances local expenditures using federal taxes. – 2nd spillover: federal fiscal authority • Monetary authority sets the nominal rate. – 3rd spillover: currency union

17 / 28

Households

( Vi (x, b, φ) = max 0

c,b ,h

)

1−θ

(1 − h) c 1−σ +ψ 1−σ 1−θ

+ β

X

Γx,x 0 Vi (x 0 , b 0 , φ0 )

x0

s.t. c + (1 + πi,t+1 )b 0 = wi si xh − T (wi si zh) + (1 + Rt−1 + κI[b<0] )b + θ(x)Di b0 ≥ b

where 0 2 − log x 0 = ρ log x + η 0 , with η ∼ iid N(0, ση ). − s is regional productivity (constant)

18 / 28

Final good firm Firm i’s preferences for local and foreign inputs are given by γii 0 ∀i 0 . " Yi =

N X

γ

1  ii 0

Z x j

i 0 =1

−1  ii 0 j

 # −1

Firm i’s demand for input j located at i 0 is:  xii 0 j = γii 0

pi 0 j Pi

− Yi

Price aggregate is " Pi =

N X i 0 =1

Z γii 0 j

1 # 1−

pi1− 0j

19 / 28

Intermediate good firms The total demand for intermediate good firm j located at region i will be yij =

N X

xi 0 ij

i 0 =1

Sticky price model: max ∗ pi,j,t

∞ X

((1 − λ)β)

s



∗ pi,j,t+s yi,j,t+s − Wi,t+s Li,j,t+s



s=0

∗ pijt  = Pit −1

PN



 1+ 0 0 0 0 Γi it+1 i 0 =1 γi i Qi 0 it wit Yi t + (1 − λ)β(1 + πi t+1 ) PN   0 0 0 0 i 0 =1 γi i Qi 0 it [Yi t + (1 − λ)β(1 + πi t+1 ) ∆i t+1 ]



with Γi 0 ,i,t = wi,t Yi 0 ,t + (1 − λ)β(1 + πi 0 ,t+1 )1+ Γi 0 ,i,t+1 ∆i 0 ,i,t = Yi 0 ,t + (1 − λ)β(1 + πi 0 ,t+1 )ε ∆i 0 ,i,t+1 Yi 0 ,t = Ci 0 ,t + Gi 0 ,t 20 / 28

Government and Monetary Authority The government buys final goods from every region Gi = {G1 , ..., GN }. • finances this expenditure using labor income taxes

¯ • issues government bonds b. Government budget constraint: Z X X XZ (1 + πi,t+1 ) bi0 − (1 + R)b¯ = Gi − T (wi si xh) i

φ

i

i

φ

The monetary authority sets the nominal rates based on a simple Taylor rule Rt = Rss + φˆ πt Equilibrium

21 / 28

Calibration

Parameter Risk Aversion Discount Factor Labor supply elasticity Disutility of labor Persistence of x Variance of innovation to x Variance of innovation to s Tax parameter Elasticity of substitution Price reset probability Dividend ownership Stock of liquid assets Borrowing limit Borrowing spread Taylor rule coefficient

Notation σ β 1/θ ψ ρ ση2 σs2 τ  λ θ B¯ b κ φ

Value

Target / Source

1 – 0.96 Nominal rate=2% 0.5 EFK-2013 0.3 Hours worked=40% 0.92 PSID 4.0% PSID 2.5% IRS 0.27 G/Y=20% 6 NS-2014 0.3 CER-2011 – Risky asset ownership (SCF) 1.2× Income SCF 0.25× Labor Income KMV-2016 0.17 19%: Net worth < 0 1.5 –

22 / 28

Cross sectional Results Figure: MPC heterogeneity Statistic

Households with b < 0 Liquid Assets/Income Mean 25th percent. 50th percent. 75th percent. 90th percent. 99th percent. Liquid Assets Gini

SCF Model (1998-2007) 19.2%

20.7%

1.20 0.01% 0.04 0.32 1.68 18.8 0.93

1.21 0.98% 0.17 1.20 3.72 12.2 0.77

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

[ Average MPC= 0.16 ]

23 / 28

Percentage Deviation from Steady-State

Local government spending shock

Government Spending

Taxes 0.014

Region 1 Region 2

0.014

0.01 0.006

0.006

0.002 -0.002

-0.002

-0.006 -0.01

-0.01 0

2

4

6

8

10 12 14 16 18 20

0

2

4

6

8

10 12 14 16 18 20

24 / 28

Percentage Deviation from Steady-State

Local government spending shock

×10 -3Inflation and nominal interest rate

3

Inflation R1 Inflation R2 Nominal Rate

2

×10 -3

Real Wage

9 7 5 3

1 1 -1

0

-3 -1

-5 0

2

4

6

8

10 12 14 16 18 20

0

2

4

6

8

10 12 14 16 18 20

24 / 28

Percentage Deviation from Steady-State

Local government spending shock

4

×10 -3

×10 -3

Consumption

Bond Holdings

2.6 3 1.8

2

1

1

0.2

0 -1

-0.6

-2

-1.4

-3

-2.2

-4

-3 0

2

4

6

8

10 12 14 16 18 20

0

2

4

6

8

10 12 14 16 18 20

24 / 28

Consumption Change– by Skill Group

Change from Steady-State

3.5

×10 -4

Region 1

5.5 4.5 3.5 2.5 1.5 0.5 -0.5 -1.5 -2.5 -3.5 -4.5 -5.5

Household Income after Taxes Consumption Asset Holdings

2.5 1.5 0.5 -0.5 -1.5 -2.5 -3.5 1

2

3

4

Skill Level

5

6

7

×10 -4

1

Region 2

2

3

4

5

6

7

Skill Level

25 / 28

Consumption Change– Decomposition

Consumption (% Change)

2

×10 -3

Region 1

1

1

0

0

-1

-1

-2

-2

×10 -3

Region 2

Benchmark Wages Inflation Nominal Rate Dividends Taxes

-3 0

2

4

6

8

10 12 14 16 18 20

0

2

4

6

8

10 12 14 16 18 20

26 / 28

Quantitative Results

Multiplier

Local

Local

Data

Consumption Wage Inflation

0.20 1.01 0.00

Aggregate Model

Empirical Evidence

0.29 2.56 0.66

-0.72 0.61 0.60

27 / 28

Quantitative Results-Specifications

Taylor rule coeff.

φ = 1.5

φ = 0.0

φ = 0.0

Reset Probability

λ = 0.3

λ = 0.3

λ = 0.0

Multiplier

(L)

(A)

Data

Consumption Wage Inflation

IRF Specification 2

0.20 1.01 0.00

(L)

(A)

(L)

(A)

0.15 2.29 0.00

0.10 3.47 0.00

Model

0.29 2.56 0.66

–0.72 0.61 0.60

0.14 2.43 0.16

0.50 4.42 0.47

IRF Specification 3

28 / 28

Thank you!

28 / 28

Placebo regressions We randomly re-assign government spending of a county to another county within the same income group/state

Spending Category

Retail Spending

Auto Spending

(Nielsen)

(Equifax)

OLS Randomly Assigned Rec. Act Spending

IV

-0.001 -0.002 (0.002) (0.002)

OLS

IV

0.001 -0.005 (0.002) (0.004)

County Controls/State F.E.

Yes

Yes

Yes

Yes

# Counties

524

524

2990

2990

Back

28 / 28

Elasticities broken down by income group Spending Category

Retail Spending

Auto Spending

(Nielsen)

(Equifax)

OLS Recovery Act Spending Rec. Act Spending × Low Income Rec. Act Spending × Medium Income Rec. Act Spending × High Income

IV

0.027 0.017 (0.018) (0.020) —

—

0.005 0.005 (0.005) (0.005) -0.007 -0.007 (0.005) (0.005)

OLS ∗∗∗

IV

0.048 (0.015)

0.056∗∗∗ (0.016)

—

—

0.011 (0.007) 0.010 (0.005)

0.012∗ (0.006) 0.011∗∗ (0.005)

County Controls/State F.E.

Yes

Yes

Yes

Yes

# Counties

525

525

2999

2930

Back

28 / 28

Percentage Deviation from Steady-State

Percentage Deviation from Steady-State

Results φ = 0.0, λ = 0.3 ×10 -3Inflation and nominal interest rate

3

×10 -3

2

Real Wage

9

Inflation R1 Inflation R2 Nominal Rate

7 5 3

1 1 -1

0

-3 -1

-5 0

4

2

4

6

×10 -3

8

10

12

14

16

18

20

0

2

4

6

×10 -3

Consumption

8

10

12

14

16

18

20

14

16

18

20

Bond Holdings

2.6 3 1.8

2

1

1

0.2

0 -1

-0.6

-2

-1.4 -2.2

-3 -4

-3 0

2

4

6

8

10

12

14

16

18

20

0

2

4

6

8

10

12

Back

28 / 28

Percentage Deviation from Steady-State

Percentage Deviation from Steady-State

Results φ = 0.0, λ = 0.0 ×10 -3Inflation and nominal interest rate

3

×10 -3

2

Real Wage

9

Inflation R1 Inflation R2 Nominal Rate

7 5 3

1 1 -1

0

-3 -1

-5 0

4

2

4

6

×10 -3

8

10

12

14

16

18

20

0

2

4

6

×10 -3

Consumption

8

10

12

14

16

18

20

14

16

18

20

Bond Holdings

2.6 3 1.8

2

1

1

0.2

0 -1

-0.6

-2

-1.4 -2.2

-3 -4

-3 0

2

4

6

8

10

12

14

16

18

20

0

2

4

6

8

10

12

Back

28 / 28

Characterizing the model

"R

Z yij = yi = j

X i0

γi 0 i

j

pij

Pi 0

#− Yi 0

"  − # X pij∗  = λ + (1 − λ)(1 + πit ) · γi 0 i Qi0 it Yi 0 Pi 0 i

• Total demand for intermediate inputs of region i is a weighted sum of final goods Y of all regions. • If demand for final good Yi0 increases then yi will increase depending on the preference of i 0 for i’s inputs γi 0 i and also on the relative price of final good Qi0 ,i . • This will affect local inflation πi and real wage wi .

Back

28 / 28

Equilibrium-Transition N 0 N N d N N We are looking to solve for {Ci,t }N 1 , {Li,t }1 , {Bi,t }1 , {Yi,t }1 , {yi,t }1 , {wi,t }1 ,

{

pi,j,t Pi,t

N }N 1 , {πi,t }1 , {Qii 0 t

N ∀i, i 0 } pairs, {Di,t }N ˆt , and τt for 1 , {φi,t }1 , Rt , π

t = {T , ∞} where T is the time of the policy change. A total of 11 × N + 3 equations. These are:

1-3) {Ci , Lsi , Bi0 }

all satisfy the households’ problem.

4) Final good i equals local consumption by households and the government: Yi = Ci + Gi ∀i = 1, N 5) Regional GDP {y1d , ..yNd } is given by "   # X pij∗ − γi 0 i Qi0 it Yi 0 ∀i = 1, N + (1 − λ)(1 + πit ) · yid = λ Pi 0 i

6) The real wage is set to equalize labor demand and supply yid = Lsi ∀i = 1, N

28 / 28

Equilibrium-Transition (cc’ed) 7) The reset price

pij∗ Pi

satisfies   PN 1+  0 0 0 Γi 0 it+1 i 0 =1 γi i Qi 0 it wit Yi t + (1 − λ)β(1 + πi t+1 )

∗ pijt  = P Pit  − 1 Ni0 =1 γi 0 i Qi0 it [Yi 0 t + (1 − λ)β(1 + πi 0 t+1 ) ∆i 0 t+1 ] ∀i = 1, N

8) The inflation rates πi,t satisfy # "   X pi∗0 j 1− 1− 1− −1 Qi 0 i + (1 − λ)(1 + πi 0 t ) 1= γii 0 Qi 0 i ∀i = 1, N λ Pi 0 9) Relative final good price pairs Qii 0 t satisfy Qi 0 it Qi 0 i(t+1)

=[

Cit σ Ci 0 t+1 σ ] ·[ ] Ci 0 t Cit+1

10) Dividends are given by "   # X pij∗ 1− Di = λ + (1 − λ)(1 + πit )−1 · [γi 0 i Qi0 i Yi 0 − wi Li ] ∀i = 1, N Pi 0 i

28 / 28

Equilibrium-Transition (cc’ed)

11) The stationary regional measures φi,t evolve based on the policy functions and the transition matrices described in the model. 12) Government budget clears X X X X X µi (1 + πi,t+1 )Bi0 − (1 + R)( µBi ) = µi G i − µi Ti − µi Fi,ss i

i

i

i

i

13) Interest rate is given by a standard Taylor rule: Rt = Rss + φˆ πt PN 14) National inflation rate is given by: π ˆt = i=1 wi πi,t , where the weights are the calculated based on the relative economic size of each region Yi,ss . Back

28 / 28

Percentage Change in State-Level Retail Spending -.05 .03 .11 .19 .27 .35

Recovery Act and retail spending, state-level

6

Spending Category

6.5 7 State-Level Recovery Act Spending (in log)

Back

Recovery Act Spending Partial F stat. State Controls/Census F.E. # States

7.5

Retail Spending (Nielsen, Store-level) OLS IV OLS IV 0.00 0.134 0.103∗ 0.273∗∗ (0.057) (0.160) (0.052) (0.108) — 132.4 — 28.3 No No Yes Yes 46 46 46 46

28 / 28

-.1

.02

County-Level PCE Inflation Rate .03 .04 .05

Percentage Change in Zip-Level Wages -.05 0 .05 .1

.06

Wages and Inflation

5

6 7 County-Level Recovery Act Spending (in log)

Category

Recovery Act Spending Partial F stat. County Controls/State F.E. # Zip codes

8

5

6 7 8 County-Level Recovery Act Spending (in log)

Wages (IRS)

Inflation (BLS)

OLS 0.025∗∗∗ (0.007)

IV 0.020∗∗ (0.008)

OLS -0.000 (0.001)

IV -0.002 (0.002)

— Yes 14602

272.7 Yes 14602

— Yes 1116

166.5 Yes 1111

Back 28 / 28