FIRM LEVERAGE, CONSUMER DEMAND, AND EMPLOYMENT …hmueller/papers/GR.pdf · FIRM LEVERAGE, CONSUMER...

FIRM LEVERAGE, CONSUMER DEMAND, ANDEMPLOYMENT LOSSES DURING THE GREAT RECESSION*

XAVIER GIROUD AND HOLGER M. MUELLER

This article argues that firms’ balance sheets were instrumental in the trans-mission of consumer demand shocks during the Great Recession. Using micro-leveldata from the U.S. Census Bureau, we find that establishments of more highlylevered firms experienced significantly larger employment losses in response todeclines in local consumer demand. These results are not driven by firms beingless productive, having expanded too much prior to the Great Recession, or be-ing generally more sensitive to fluctuations in either aggregate employment orhouse prices. Likewise, at the county level, we find that counties with more highlylevered firms experienced significantly larger declines in employment in responseto local consumer demand shocks. Accordingly, firms’ balance sheets also matterfor aggregate employment. Our results suggest a possible role for employmentpolicies that target firms directly besides conventional stimulus. JEL Codes: E24,E32, G32, J21, J23, R31.

I. INTRODUCTION

The collapse in house prices during the Great Recessioncaused a sharp drop in consumer demand by households (Mian,Rao, and Sufi 2013). This drop in consumer demand, in turn,had severe consequences for employment: across U.S. counties,those with larger declines in housing net worth experiencedsignificantly larger declines in employment, especially in thenontradable sector (Mian and Sufi 2014a).

*We are grateful to Andrei Shleifer (editor), Larry Katz (coeditor), and threeanonymous referees for helpful comments and suggestions. We also thank NickBloom, Itamar Drechsler, Mathias Drehmann, Andrea Eisfeldt, Mark Gertler,Ed Glaeser, Bob Hall, Theresa Kuchler, Atif Mian, Jordan Nickerson, Stijn vanNieuwerburgh, Jonathan Parker, Andrew Patton, Thomas Philippon, Alexi Savov,Phillip Schnabl, Antoinette Schoar, Rui Silva, Johannes Stroebel, Amir Sufi, andseminar participants at MIT, NYU, Stanford, UBC, UNC, UT Austin, New YorkFed, Purdue, Nova, NBER EF&G Summer Institute, NBER CF Meeting, LBSSummer Finance Symposium, ITAM Finance Conference, and CEPR/Universityof St. Gallen Workshop on Household Finance and Economic Stability. ManuelAdelino, Wei Jiang, and Albert Saiz kindly provided us with data. Vadim Elenevprovided excellent research assistance. Any opinions and conclusions expressedherein are those of the authors and do not necessarily represent the views of theU.S. Census Bureau. All results have been reviewed to ensure that no confidentialinformation is disclosed.C© The Author(s) 2016. Published by Oxford University Press, on behalf of Presidentand Fellows of Harvard College. All rights reserved. For Permissions, please email:[email protected] Quarterly Journal of Economics (2017), 271–316. doi:10.1093/qje/qjw035.Advance Access publication on October 13, 2016.

271

mailto:[email protected]

272 QUARTERLY JOURNAL OF ECONOMICS

What is conspicuously absent from this causal chain is anyrole for firms. After all, households do not lay off workers. Firmsdo. To investigate the role of firms in the transmission of con-sumer demand shocks during the Great Recession, we construct aunique data set that combines employment data at the establish-ment level from the U.S. Census Bureau’s Longitudinal BusinessDatabase (LBD) with balance sheet and income statement dataat the firm level from Compustat and house price data at thezip code and county level from Zillow. Hence, our sample consistsof individual establishments—for example, retail stores, super-markets, or restaurants—that are matched to house prices in theestablishment’s zip code or county.

Our results suggest that firm balance sheets played a crucialrole in the transmission of consumer demand shocks during theGreat Recession. This is noteworthy because both academic re-search and public policy have hitherto primarily focused on eitherhousehold or financial intermediary balance sheets.1 In partic-ular, our results show that establishments of firms with higherleverage at the onset of the Great Recession experienced signif-icantly larger employment losses in response to declines in localconsumer demand during the Great Recession. The magnitudeof this leverage effect is large. Imagine two establishments, onewhose parent firm lies at the 90th percentile of the leverage distri-bution and another whose parent firm lies at the 10th percentile ofthe leverage distribution. Our estimates imply that the former es-tablishment exhibits a three times larger elasticity of employmentwith respect to local house prices. Importantly, the correlation be-tween firm leverage and changes in house prices during the GreatRecession is virtually 0. Thus, establishments of low- and high-leverage firms face similar local consumer demand shocks—theymerely react differently to these shocks.

The granularity of our data allows us to include a wide ar-ray of fixed effects in our regressions. Our tightest specification

1. For research focusing on the role of household balance sheets in the GreatRecession, see, for example, Hall (2011), Eggertson and Krugman (2012), Mian,Rao, and Sufi (2013), Mian and Sufi (2014a), Guerrieri and Lorenzoni (2015),and Midrigan and Philippon (2016). For research focusing on the role of finan-cial intermediary balance sheets, and “lender health” more generally, see Gertlerand Kyotaki (2011), He and Krishnamurthy (2013), Brunnermeier and Sannikov(2014), Chodorow-Reich (2014), and Moreira and Savov (2016). A notable excep-tion is Gilchrist et al. (2016), who show that firms with weak balance sheets raiseprices during the Great Recession, which may help explain why the U.S. economyexperienced only a mild disinflation during that period.

FIRM LEVERAGE, CONSUMER DEMAND, EMPLOYMENT LOSSES273

includes both firm and zip code × industry fixed effects. Hence,accounting for the possibility that low- and high-leverage firmsexperience differential employment losses for reasons unrelatedto changes in consumer demand, our empirical setting allows us tocompare establishments in the same zip code and industry, wheresome establishments belong to low-leverage firms and others be-long to high-leverage firms. Our establishment-level results arebased on more than a quarter million observations and are thusprecisely estimated.

We also examine whether firms make adjustments at theextensive margin. Similar to our employment results, establish-ments of more highly levered firms are significantly more likelyto be closed down in response to declines in local consumer de-mand. Furthermore, and in line with prior research, we findno significant correlation between changes in house prices andchanges in establishment-level employment in the tradable sec-tor. By contrast, we find positive and significant correlations in thenontradable and “other” sector—that is, industries that are nei-ther tradable nor nontradable. Importantly, in both sectors, thiscorrelation is significantly stronger among establishments of morehighly levered firms.

Our results are consistent with financial constraints impair-ing firms’ ability to engage in labor hoarding. The idea behindlabor hoarding is that firms facing a temporary (e.g., cyclical)decline in demand retain more workers than technically neces-sary so as to economize on the costs of firing, hiring, and train-ing workers. Labor hoarding is costly, however. Effectively firmsmust (temporarily) subsidize workers’ wages. Hence, firms withlittle financial slack face a genuine trade-off between long-runoptimization—saving on firing, hiring, and training costs—andshort-run liquidity needs. Our results suggest that firms withweaker balance sheets—and tighter financial constraints—aremore apt to respond to this trade-off by engaging in less laborhoarding.

In our sample, more highly levered firms indeed appear to bemore financially constrained based on various measures. But dothey also act like financially constrained firms in the Great Reces-sion? To address this question, we turn to firm-level regressions.Indeed, we find that more highly levered firms are less apt (orable) to raise additional short- and long-term debt in response toa decline in local consumer demand. As a consequence, they expe-rience more layoffs, are more likely to close down establishments,


and cut back more on investment. Altogether, our results suggestthat firms with higher leverage not only appear to be more finan-cially constrained but also act like financially constrained firmsduring the Great Recession.

An important concern is that more highly levered firms re-spond more strongly to local consumer demand shocks not be-cause they are more financially constrained but because of otherreasons unrelated to financial constraints. Although we cannotrule out this possibility in general, we can address specific al-ternative stories. For instance, more highly levered firms maybe less productive, have expanded too much prior to the GreatRecession, or have more active investors, such as private eq-uity funds and activist hedge funds. Or high-leverage firms maysimply be “high-beta” firms that are generally more sensitive toeither aggregate employment or house prices—that is, for rea-sons unrelated to financial constraints. We find little evidence insupport of these alternative stories. Furthermore, there is vir-tually no correlation between firm leverage and either housingsupply elasticity or changes in house prices during the Great Re-cession. Hence, more highly levered firms do not respond morestrongly to consumer demand shocks because they are located inregions with stronger shocks. In fact, given our fixed-effect speci-fication, we can rule out any alternative story in which low- andhigh-leverage firms differ along either geographical or industrydimensions.

In general equilibrium, output and workers may shift fromhigh- to low-leverage firms. In an economy without frictions, thiscould imply that aggregate employment changes little or perhapsnot at all. To empirically investigate whether the distribution offirm leverage also matters in the aggregate, we turn to county-level regressions. Imagine two counties, one with a smaller shareof high-leverage firms and the other with a larger share. Supposethat both counties exhibit a similar drop in house prices. If ourprevious results also hold in the aggregate, then the more highlylevered county should experience a larger decline in county-levelemployment. By contrast, if the distribution of firm leverage doesnot matter in the aggregate, then both counties should experi-ence similar declines in county-level employment, irrespective ofthe level of county leverage. Regardless of whether we considercounty-level employment by all firms in our sample or by all firmsin the LBD, we find that more highly levered counties exhibit sig-nificantly larger employment losses in response to local consumer


demand shocks. Thus, our results are not undone by general equi-librium effects.

As in our establishment-level analysis, we also examine al-ternative channels at the county level. We find no evidence thatmore highly levered counties respond more strongly to consumerdemand shocks during the Great Recession because their firmsare less productive, have expanded too much in prior years, orhave more activist investors, such as private equity funds andactivist hedge funds, or because employment in these counties isgenerally more sensitive to either aggregate employment or houseprices.

We conclude with a discussion of policy implications. That fi-nancial constraints may impair firms’ ability to engage in laborhoarding suggests it may be useful to think about policies thattarget firms directly besides conventional stimulus. To this end,we discuss the case of Germany, which has seen virtually no risein unemployment despite being hit hard by the global recession of2008–2009. Many commentators attribute this resilience to mas-sive labor hoarding, which is heavily subsidized in Germany. Acentral pillar of German labor hoarding is the system of “short-time work” programs encouraging firms to adjust labor demandthrough hours reductions rather than layoffs. Although a similarsystem also exists in many U.S. states (“work-sharing” programs),take-up rates have been extremely low due to burdensome filingprocesses, program rigidity, and financial disincentives for em-ployers and workers.

In seminal work, Mian and Sufi (2011) and Mian, Rao, andSufi (2013) document that rising house prices in the years priorto the Great Recession led to the build-up of household leverage,causing a sharp drop in consumer demand as house prices plum-met during the Great Recession. Mian and Sufi (2014a) examinethe consequences of these consumer demand shocks for aggregateemployment at the county level.2 Our focus is at the individualestablishment level. In particular, we find that establishments ofmore highly levered firms exhibit significantly larger employmentlosses in response to declines in local consumer demand duringthe Great Recession.

The notion that firm balance sheets play an important role inthe transmission of business cycle shocks goes back to Bernanke

2. Charles, Hurst, and Notowidigdo (2015) and Adelino, Schoar, and Severino(2015) examine the role of rising house prices for employment growth in the yearsleading up to the Great Recession.


and Gertler (1989), Kiyotaki and Moore (1997), and Bernanke,Gertler, and Gilchrist (1999). Unlike a “standard” financial accel-erator model, however, our focus is not on aggregate shocks tofirms’ net worth but rather on the interaction between heteroge-neous demand shocks and firm balance sheets. Caggese and Perez(2016) model precisely this interaction in a dynamic general equi-librium model with heterogeneous firms and households subjectto financial and labor market frictions. When calibrating theirmodel to U.S. data, they find interaction effects which, as theyconclude, are in line with those found in our article. Aghion et al.(2015) also explore the role of firm heterogeneity during the GreatRecession. Using firm-level data from Organisation for EconomicCo-operation and Development (OECD) countries, they find thatdecentralized firms fare significantly better than their central-ized counterparts, especially in industries that were hit hard bythe Great Recession.

The rest of this article is organized as follows. Section II de-scribes the data, variables, and summary statistics. Section IIIexamines the interplay between consumer demand shocks, firmbalance sheets, and employment at the establishment level.Section IV discusses financial constraints and labor hoarding.Section V explores alternative channels. Section VI considers ag-gregate employment at the county level. Section VII discussespolicy implications. Section VIII concludes.

II. DATA, VARIABLES, AND SUMMARY STATISTICS

We construct a unique data set that combines employmentdata at the establishment level with balance sheet and incomestatement data at the firm level and house price data at the zipcode and county level.

The establishment-level data are provided by the U.S. CensusBureau’s LBD. An establishment is a “single physical locationwhere business is conducted” (Jarmin and Miranda 2002, p. 5), forexample, a retail store, supermarket, restaurant, warehouse, ormanufacturing plant. The LBD covers all business establishmentsin the United States with at least one paid employee.

The firm-level data are from Compustat. We exclude financialfirms (SIC 60–69), utilities (SIC 49), and firms with missing finan-cial data between 2002 and 2009. We match the remaining firmsto establishments in the LBD using the Compustat-SSEL bridgemaintained by the U.S. Census Bureau. Because this bridge ends


in 2005, we extend the match to 2009 using employer name and IDnumber (EIN) following the procedure described in McCue (2003).This leaves us with 327,500 establishments with nonmissing em-ployment data from 2007 to 2009.3

The house price data are from Zillow.4 Of the 327,500 estab-lishments, we are able to match 227,600 establishments to zipcode-level house prices and 57,200 establishments to county-levelhouse prices, leaving us with a final sample of 284,800 establish-ments for which we have both firm-level data and house pricedata.5

Our main analysis is at the establishment level. We regressthe percentage change in establishment-level employment be-tween 2007 and 2009, �Log(Emp)07−09, on the percentage changein house prices in the establishment’s zip code or county (ifthe zip code information is missing) between 2006 and 2009,�Log(HP)06−09, the level of firm leverage associated with the es-tablishment’s parent firm in 2006, Leverage06, and the interactionterm �Log(HP)06−09 × Leverage06. Our main focus is on the inter-action term. Leverage is defined as the ratio of the sum of debt incurrent liabilities and long-term debt to total assets (from Com-pustat) and is winsorized between 0 and 1. In robustness checks,we also use other measures of firms’ balance sheet strength. Al-though our main specification includes industry fixed effects, someof our specifications also include firm, zip code, or zip code × in-dustry fixed effects. All regressions are weighted by the size ofestablishments, that is, their number of employees. Standard er-rors are clustered at both the state and firm level.

Changes in house prices from 2006 to 2009 based on Zillowdata are highly correlated with the “housing net worth shock” inMian, Rao and Sufi (2013) and Mian and Sufi (2014a), �HousingNet Worth, 2006–2009. The correlation at the metropolitan statis-tical area (MSA) level is 86.3%. Other studies, like Adelino, Schoar,

3. All sample sizes are rounded to the nearest hundred following disclosureguidelines by the U.S. Census Bureau.

4. For the period from 2006 to 2009, the Zillow Home Value Index(ZHVI) is available for 12,102 zip codes and 1,048 counties. See http://www.zillow.com/research/data for an overview of the ZHVI methodology and a com-parison with the S&P/Case-Shiller Home Price Index.

5. Our results are similar if we use only the 227,600 establishments for whichwe have zip code-level house prices or if we use the full sample of 327,500 establish-ments by matching the remaining 327,500 − 284,800 = 42,700 establishments tostate-level house prices constructed as population-weighted averages of availablezip code-level house prices. See Table 1 of the Online Appendix.

http://www.zillow.com/research/data

http://www.zillow.com/research/data

https://academic.oup.com/qje


and Severino (2015) and Charles, Hurst, and Notowidigdo (2015),use house price data from the Federal Housing Finance Agency(FHFA). The correlation at the MSA level between changes inhouse prices from 2006 to 2009 based on Zillow data and FHFAdata is 96.4%.

Table I, Panel A provides summary statistics at the estab-lishment level. The first and second columns show the meanand standard deviation, respectively. The third column showsthe correlation with Leverage06. The last column shows thep-value of this correlation. As can be seen, Leverage06 is uncor-related with changes in house prices between 2006 and 2009 andhousing supply elasticity. Thus, establishments of low- and high-leverage firms face similar consumer demand shocks during theGreat Recession—they merely react differently to these shocks.Interestingly, establishments of more highly levered firms are un-derrepresented in the nontradable sector and overrepresented inthe “other” sector—that is, industries that are neither tradablenor nontradable. This is not a major concern, however. First, weperform separate analyses for each industry sector and find sim-ilar results in the nontradable and “other” sector. Second, all ourregressions include industry fixed effects.

Panel B provides firm-level summary statistics in 2006, atthe onset of the Great Recession. As can be seen, more highlylevered firms are less productive—they have a lower return on as-sets (ROA), lower net profit margin (NPM), and lower total factorproductivity (TFP). Moreover, and perhaps not surprising, morehighly levered firms are more financially constrained accordingto the financial constraints indices of Kaplan and Zingales (1997)and Whited and Wu (2006).

Panel C includes the same firm-level variables as Panel B.However, instead of showing their levels in 2006, it shows theirchanges between 2002 and 2006. Three results stand out. First,firms with higher leverage expand more in the years prior to theGreat Recession. This holds irrespective of whether we considerestablishment, employment, or asset growth. Second, firms withhigher leverage exhibit lower productivity growth between 2002and 2006, which may explain the negative correlation betweenleverage and productivity in Panel B. Third, firms with higherleverage experience an increase in leverage along with a tighten-ing of financial constraints in the years before the Great Reces-sion. Although this latter result is not surprising, the magnitudeis large: the correlation between Leverage06 and the change in

FIRM LEVERAGE, CONSUMER DEMAND, EMPLOYMENT LOSSES279T

AB

LE

IS

UM

MA

RY

ST

AT

IST

ICS

Cor

rela

tion

wit

hp-

valu

eof

Mea

nS

td.d

ev.

Lev

erag

e 06

corr

elat

ion

Pan

elA

:Est

abli

shm

ent

leve

l(N

=28

4,80

0)E

mpl

oyee

s 06

3963

−0.0

28.2

83�

Log

(Em

p)07

−09

−8.2

24.2

−0.0

47**

.020

�L

og(H

P) 0

6−09

−14.

516

.10.

005

.718

Hou

sin

gsu

pply

elas

tici

ty1.

799

0.92

70.

011

.345

Cen

sus

regi

ons

Nor

thea

st0.

170.

380.

004

.801

Mid

wes

t0.

210.

41−0

.006

.610

Sou

th0.

380.

490.

002

.894

Wes

t0.

240.

42−0

.000

.978

Indu

stry

sect

ors

Tra

dabl

e0.

030.

180.

001

.954

Non

trad

able

0.44

0.5

−0.1

46**

.014

Oth

er0.

530.

50.

145*

*.0

13P

anel

B:F

irm

leve

l200

6(N

=2,

800)

Est

abli

shm

ents

0610

145

1−0

.015

.495

Em

ploy

ees 0

64,

005

16,3

84−0

.008

.191

Ass

ets 0

63,

040

18,5

15−0

.003

.655

RO

A06

0.04

50.

177

−0.0

73**

*.0

03N

PM

060.

024

0.28

−0.0

41**

.032

TF

P06

−0.0

020.

599

−0.0

83**

*.0

04L

ever

age 0

60.

227

0.25

31.

000*

**.0

00W

W06

−0.2

510.

135

0.18

9***

.000

KZ

06−4

.067

44.2

950.

259*

**.0

00

280 QUARTERLY JOURNAL OF ECONOMICST

AB

LE

I( C

ON

TIN

UE

D)

Cor

rela

tion

wit

hp-

valu

eof

Mea

nS

td.d

ev.

Lev

erag

e 06

corr

elat

ion

Pan

elC

:Fir

mle

vel2

002–

2006

(N=

2,80

0)�

Est

abli

shm

ents

02−0

64.

410

.20.

089*

**.0

00�

Log

(Em

p)02

−06

0.05

20.

093

0.04

8***

.009

�L

og(A

sset

s)02

−06

0.11

00.

133

0.08

7***

.000

�R

OA

02−0

60.

022

0.12

7−0

.061

***

.003

�N

PM

02−0

60.

020

0.22

5−0

.032

**.0

15�

TF

P02

−06

−0.0

010.

569

−0.0

17.6

49�

Lev

erag

e 02−

06−0

.023

0.15

30.

379*

**.0

00�

WW

02−0

6−0

.006

0.08

0.08

5***

.000

�K

Z02

−06

−0.3

7049

.633

0.18

8***

.000

Pan

elD

:10t

han

d90

thpe

rcen

tile

sof

�L

og(H

P) 0

6−09

and

leve

rage

0610

th90

th�

Log

(HP

) 06−

09−0

.402

0.03

5L

ever

age 0

60.

000

0.56

8

Not

es.P

anel

Apr

ovid

essu

mm

ary

stat

isti

csat

the

esta

blis

hm

ent

leve

l.L

ever

age 0

6is

the

rati

oof

the

sum

ofde

btin

curr

ent

liab

ilit

ies

and

lon

g-te

rmde

btto

tota

lass

ets

asso

ciat

edw

ith

the

esta

blis

hm

ent’s

pare

nt

firm

in20

06.�

Log

(Em

p)07

−09

isth

epe

rcen

tage

chan

gein

esta

blis

hm

ent-

leve

lem

ploy

men

tfr

om20

07to

2009

.�

Log

(HP

) 06−

09is

the

perc

enta

gech

ange

inh

ouse

pric

esin

the

esta

blis

hm

ent’s

zip

code

orco

un

ty(i

fth

ezi

pco

dein

form

atio

nis

mis

sin

g)fr

om20

06to

2009

.H

ousi

ng

Su

pply

Ela

stic

ity

isde

scri

bed

inS

aiz

(201

0).

Tra

dabl

ean

dn

ontr

adaa

ble

indu

stri

esar

ede

scri

bed

inM

ian

and

Su

fi(2

014a

).“O

ther

”in

dust

ries

are

thos

eth

atar

en

eith

ertr

adab

len

orn

ontr

adab

le.

Pan

els

Ban

dC

prov

ide

sum

mar

yst

atis

tics

atth

efi

rmle

vel.

Ass

ets

isth

ebo

okva

lue

ofto

tala

sset

s.R

OA

(ret

urn

onas

sets

)is

the

rati

oof

oper

atin

gin

com

ebe

fore

depr

ecia

tion

toto

tala

sset

s.N

PM

(net

profi

tm

argi

n)

isth

era

tio

ofop

erat

ing

inco

me

befo

rede

prec

iati

onto

sale

s.T

FP

(tot

alfa

ctor

prod

uct

ivit

y)is

the

resi

dual

from

are

gres

sion

oflo

g(sa

les)

onlo

g(em

ploy

ees)

and

log(

PP

&E

)ac

ross

all

Com

pust

atfi

rms

inth

esa

me

two-

digi

tS

ICin

dust

ry.W

Wis

the

fin

anci

alco

nst

rain

tsin

dex

ofW

hit

edan

dW

u(2

006)

.KZ

isth

efi

nan

cial

con

stra

ints

inde

xof

Kap

lan

and

Zin

gale

s(1

997)

.Pan

elD

show

sth

e10

than

d90

thpe

rcen

tile

sof

�L

og(H

P) 0

6−09

and

Lev

erag

e 06,r

espe

ctiv

ely,

use

din

cou

nte

rfac

tual

anal

yses

inS

ecti

onII

I.A

.**,

and

***

den

otes

sign

ifica

nce

atth

e5%

,an

d1%

leve

ls,r

espe

ctiv

ely.


firm leverage between 2002 and 2006, �Leverage02−06, is 37.9%.Hence, a substantial portion of the cross-sectional variation infirm leverage in 2006, at the onset of the Great Recession, can beexplained by changes in firm leverage in the years leading up tothe Great Recession.

Panel D shows the 10th and 90th percentiles of �Log(HP)06−09and Leverage06. We make use of these percentiles below to performcounterfactual analyses.

Last, we caution that the various correlations with Leverage06may not be independent of each other. Indeed, it is plausible thatfirms with higher leverage in 2006 increased their leverage be-tween 2002 and 2006 because they needed to fund an expansionor a deficit arising from a productivity shortfall. But this raises thepossibility that firms with higher leverage in 2006 respond morestrongly to consumer demand shocks in the Great Recession notbecause they are more financially constrained but because theyare less productive or expanded too much in the years prior tothe Great Recession. We will address these and other alternativestories below.

III. FIRM LEVERAGE, CONSUMER DEMAND, AND EMPLOYMENT

III.A. Main Results

To obtain a visual impression, Figure I plots the percentagechange in establishment-level employment between 2007 and2009, �Log(Emp)07−09, against the percentage change in houseprices in the establishment’s zip code or county (if the zip codeinformation is missing) between 2006 and 2009, �Log(HP)06−09,for different quartiles of firm leverage. For each percentile of�Log(HP)06−09, the plot shows the mean values of �Log(HP)06−09and �Log(Emp)07−09, respectively. In Panel A, which representsthe lowest leverage quartile, there is a positive albeit weakrelationship between changes in house prices and changes inestablishment-level employment, as is illustrated by the solidtrend line. In Panels B to D, this relationship becomes succes-sively stronger. In Panel D, which represents the highest leveragequartile, the elasticity of establishment-level employment withrespect to house prices is 0.096, which is more than four timeslarger than the elasticity in the lowest leverage quartile.

Table II confirms this visual impression using regres-sion analysis. All regressions are weighted by the size of


FIG

UR

EI

Fir

mL

ever

age,

Hou

seP

rice

s,an

dE

mpl

oym

ent

atth

eE

stab

lish

men

tL

evel

Th

efi

gure

plot

sth

epe

rcen

tage

chan

gein

esta

blis

hm

ent-

leve

lem

ploy

men

tbe

twee

n20

07an

d20

09,�

Log

(Em

p)07

−09,

agai

nst

the

perc

enta

gech

ange

inh

ouse

pric

esin

the

esta

blis

hm

ent’s

zip

code

orco

un

tybe

twee

n20

06an

d20

09,�

Log

(HP

) 06−

09,f

ordi

ffer

entq

uar

tile

sof

firm

leve

rage

.For

each

perc

enti

leof

�L

og(H

P) 0

6−09

,th

esc

atte

rplo

tde

pict

sth

em

ean

valu

esof

�L

og(H

P) 0

6−09

and

�L

og(E

mp)

07−0

9,

resp

ecti

vely

.


TA

BL

EII

FIR

ML

EV

ER

AG

E,C

ON

SU

ME

RD

EM

AN

D,A

ND

EM

PL

OY

ME

NT

�L

og(E

mp)

07−0

9(1

)(2

)(3

)(4

)(5

)(6

)(7

)

�L

og(H

P) 0

6−09

0.06

6***

0.02

90.

029

0.02

7(0

.019

)(0

.022

)(0

.019

)(0

.019

)�

Log

(HP

) 06−

09×

Lev

erag

e 06

0.11

1***

0.11

4***

0.11

3***

0.08

4**

0.07

6**

0.07

5**

(0.0

39)

(0.0

40)

(0.0

38)

(0.0

35)

(0.0

31)

(0.0

38)

Lev

erag

e 06

−0.0

28**

−0.0

32**

−0.0

20**

(0.0

14)

(0.0

15)

(0.0

09)

Indu

stry

fixe

def

fect

sN

oN

oYe

sYe

sYe

sYe

s—

Fir

mfi

xed

effe

cts

No

No

No

No

Yes

Yes

Yes

Zip

code

fixe

def

fect

sN

oN

oN

oYe

sN

oYe

s—

Zip

code

×in

dust

ryfi

xed

effe

cts

No

No

No

No

No

No

Yes

R-s

quar

ed0.

000.

000.

040.

130.

170.

250.

31O

bser

vati

ons

284,

800

284,

800

284,

800

284,

800

284,

800

284,

800

284,

800

Not

es.

Th

ede

pen

den

tva

riab

le,�

Log

(Em

p)07

−09,

isth

epe

rcen

tage

chan

gein

esta

blis

hm

ent-

leve

lem

ploy

men

tfr

om20

07to

2009

.�

Log

(HP

) 06−

09is

the

perc

enta

gech

ange

inh

ouse

pric

esin

the

esta

blis

hm

ent’s

zip

code

orco

un

ty(i

fth

ezi

pco

dein

form

atio

nis

mis

sin

g)fr

om20

06to

2009

.Lev

erag

e 06

isth

era

tio

ofth

esu

mof

debt

incu

rren

tli

abil

itie

san

dlo

ng-

term

debt

toto

tal

asse

tsas

soci

ated

wit

hth

ees

tabl

ish

men

t’spa

ren

tfi

rmin

2006

.In

dust

ryfi

xed

effe

cts

are

base

don

fou

r-di

git

NA

ICS

code

s.A

llre

gres

sion

sar

ew

eigh

ted

byes

tabl

ish

men

tsi

ze.S

tan

dard

erro

rs(i

npa

ren

thes

es)

are

clu

ster

edat

both

the

stat

ean

dfi

rmle

vel.

**,a

nd

***

den

ote

sign

ifica

nce

atth

e5%

,an

d1%

leve

ls,r

espe

ctiv

ely.


establishments. Standard errors are clustered at the state andfirm level.6 As can be seen, the average elasticity of establishment-level employment with respect to house prices is 0.066 (column(1)). To put this number into perspective, imagine two estab-lishments, one located in a zip code associated with a 10th per-centile change in house prices and another in a zip code associ-ated with a 90th percentile change in house prices. An elasticityof 0.066 implies that the former establishment experiences anadditional employment loss of 2.88 percentage points.7 Accord-ingly, changes in house prices during the Great Recession have aprofound impact on changes in employment at the establishmentlevel.

Columns (2) to (7) examine whether the elasticity ofestablishment-level employment with respect to house prices de-pends on the leverage of the establishment’s parent firm. Eachcolumn has a different set of fixed effects. Arguably, our “tight-est” specification is that in column (7). While the inclusion of firmfixed effects accounts for any unobserved heterogeneity acrossfirms, the zip code × industry fixed effects force comparison to bemade between establishments in the same zip code and four-digitNAICS industry. Note that although our sample firms are in Com-pustat, their establishments are relatively small, with an averagesize of 39 employees. Thus, accounting for the possibility thatlow- and high-leverage firms may exhibit differential job lossesfor reasons unrelated to changes in house prices, our empiricalsetting compares relatively small establishments in the same zipcode and four-digit NAICS industry, where some establishmentsbelong to low-leverage firms and others belong to high-leveragefirms.

Regardless of which fixed effects we include, we always findthat the interaction term �Log(HP)06−09 × Leverage06 is posi-tive and significant. Hence, establishments of more highly lev-ered firms exhibit significantly larger declines in employment inresponse to local consumer demand shocks. The magnitude of thisleverage effect is large. Imagine two establishments, one whoseparent firm lies at the 90th percentile and another whose par-ent firm lies at the 10th percentile of the leverage distribution.Our estimates in column (3) imply that the former establishment

6. Table 2 of the Online Appendix considers alternative clustering methods.7. See Panel D of Table I: 0.066 × (0.035 − (−0.402)) = 0.0288, where 0.035 and

−0.402 represent the 10th and 90th percentile, respectively, of �Log(HP)06−09.



exhibits a three times larger elasticity of employment with respectto house prices.8

The only fixed effect that has a noticeable impact on the co-efficient associated with the interaction term is the firm fixedeffect. Moving from columns (2) to (4) to columns (5) to (7), whichinclude firm fixed effects, the coefficient associated with the in-teraction term drops markedly, although it remains significant atthe 5% level. Note, however, that including firm fixed effects maybe “overcontrolling”—that is, it may be “controlling away” someof the very effect we are trying to document. For instance, somefirms in our sample have most of their establishments in the sameregion. As the inclusion of firm fixed effects forces comparison tobe made between different establishments within the same firm,this implies that for regionally concentrated firms, there existsonly little within-firm variation in house price changes, makingit difficult to identify the effect on employment changes. Alterna-tively, internal capital market flows may level out differences inemployment losses across establishments within the same firm. Ifthe firm’s headquarters engages in cross-subsidization, establish-ments in less affected regions may subsidize those in more affectedregions, reducing the within-firm variation in the sensitivity ofemployment with respect to local consumer demand shocks. Giventhese issues, we use column (3) as our main specification. Thisspecification has the additional advantage that it also shows thecoefficients associated with the main effects, �Log(HP)06−09 andLeverage06, respectively. That being said, the analysis in Table IIhas shown that our main results hold under various fixed-effectspecifications.

III.B. Other Measures of Firm’s Balance Sheet Strength

We obtain similar results when using other measures of firms’debt capacity or balance sheet strength. As Table III shows, allresults are similar when using either net or market leverage,debt to EBITDA, and interest coverage, all measured in 2006(columns (1) to (4)). They are also similar when using the changein leverage between 2002 and 2006 in lieu of the level of leverage

8. See Panel D of Table I: 0.029 + 0.114 × 0.568 = 0.094 versus 0.029 + 0.114 ×0.000 = 0.029, where 0.568 and 0.000 represent the 90th and 10th percentile,respectively, of Leverage06. Although these counterfactual elasticities are basedon a linear specification, they compare reasonably well to the elasticities in thehighest and lowest leverage quartiles in Figure I.


TA

BL

EII

IO

TH

ER

ME

AS

UR

ES

OF

FIR

MS’B

AL

AN

CE

SH

EE

TS

TR

EN

GT

H

�L

og(E

mp)

07−0

9N

etM

arke

tD

ebt

toIn

tere

stle

vera

ge06

leve

rage

06E

BIT

DA

06co

vera

ge06

�L

ever

age 0

2−06

WW

06K

Z06

(1)

(2)

(3)

(4)

(5)

(6)

(7)

�L

og(H

P) 0

6−09

0.02

90.

032

0.03

6*0.

040*

0.01

10.

027

0.02

9(0

.020

)(0

.019

)(0

.019

)(0

.020

)(0

.033

)(0

.023

)(0

.021

)�

Log

(HP

) 06−

09×

Deb

tca

paci

ty0.

120*

**0.

130*

**0.

012*

**0.

127*

*0.

223*

**0.

059*

**0.

003*

*(0

.041

)(0

.044

)(0

.004

)(0

.056

)(0

.070

)(0

.015

)(0

.002

)D

ebt

capa

city

−0.0

38**

−0.0

54**

*−0

.003

**−0

.063

***

−0.0

38*

−0.0

09**

−0.0

03**

*(0

.016

)(0

.018

)(0

.001

)(0

.024

)(0

.022

)(0

.004

)(0

.000

)In

dust

ryfi

xed

effe

cts

Yes

Yes

Yes

Yes

Yes

Yes

Yes

R-s

quar

ed0.

040.

040.

040.

040.

040.

040.

04O

bser

vati

ons

284,

800

284,

800

284,

800

284,

800

284,

800

284,

800

284,

800

Not

es.T

his

tabl

epr

esen

tsva

rian

tsof

the

regr

essi

ons

inT

able

IIin

wh

ich

Lev

erag

e 06

isre

plac

edby

oth

erm

easu

res

offi

rms’

debt

capa

city

orba

lan

cesh

eet

stre

ngt

h.N

etle

vera

geis

the

rati

oof

debt

incu

rren

tli

abil

itie

spl

us

lon

g-te

rmde

btm

inu

sca

shan

dsh

ort-

term

inve

stm

ents

divi

ded

byto

tala

sset

s.M

arke

tle

vera

geis

the

rati

oof

debt

incu

rren

tli

abil

itie

spl

us

lon

g-te

rmde

btdi

vide

dby

tota

las

sets

min

us

the

book

valu

eof

equ

ity

plu

sth

em

arke

tva

lue

ofeq

uit

y(s

tock

pric

em

ult

ipli

edby

the

nu

mbe

rof

shar

esou

tsta

ndi

ng)

.D

ebt

toE

BIT

DA

isth

era

tio

ofde

btin

curr

ent

liab

ilit

ies

plu

slo

ng-

term

debt

divi

ded

byop

erat

ing

inco

me

befo

rede

prec

iati

on.I

nte

rest

cove

rage

isth

era

tio

ofin

tere

stex

pen

seto

oper

atin

gin

com

eaf

ter

depr

ecia

tion

.�L

ever

age 0

2−06

isth

ech

ange

infi

rmle

vera

gefr

om20

02to

2006

.WW

and

KZ

are

the

fin

anci

alco

nst

rain

tsin

dexe

sof

Wh

ited

and

Wu

(200

6)an

dK

apla

nan

dZ

inga

les

(199

7),r

espe

ctiv

ely.

�L

ever

age,

WW

,an

dK

Zar

en

etof

thei

rre

spec

tive

min

imu

mva

lues

.In

dust

ryfi

xed

effe

cts

are

base

don

fou

r-di

git

NA

ICS

code

s.A

llre

gres

sion

sar

ew

eigh

ted

byes

tabl

ish

men

tsi

ze.S

tan

dard

erro

rs(i

npa

ren

thes

es)

are

clu

ster

edat

both

the

stat

ean

dfi

rmle

vel.

*,**

,an

d**

*de

not

esi

gnifi

can

ceat

the

10%

,5%

,an

d1%

leve

ls,

resp

ecti

vely

.


in 2006 (column (5)). As discussed previously, these two variablesare highly correlated, implying that firms with higher leverage in2006 are largely firms that increased their leverage in previousyears. Finally, our results are similar when using the financialconstraints indexes of Kaplan and Zingales (1997) and Whited andWu (2006) (columns (6) and (7)). Ultimately, all of the measuresin Table III are proxies for the strength of firms’ balance sheets.

III.C. Instrumenting House Price Changes

Unobserved heterogeneity may be driving changes in houseprices and changes in employment. We address this issue by in-strumenting changes in house prices using the housing supplyelasticity instrument from Saiz (2010). This instrument capturesgeographical and regulatory constraints to new construction. Ac-cordingly, areas with inelastic housing supply face supply con-straints due to their topography (steep hills and water bodies) aswell as local regulations. The Saiz instrument has been widelyused in the literature as a source of exogenous variation in houseprice changes (e.g., Mian and Sufi 2011, 2014a; Mian, Rao, andSufi 2013; Adelino, Schoar, and Severino 2015; Baker 2015; Bergerand Vavra 2015; Stroebel and Vavra 2015; Kaplan, Mitman, andViolante 2016).

The instrumental variables (IV) results are provided inTable 3 of the Online Appendix. Similar to other studies, we findthat housing supply elasticity is a strong predictor of changesin house prices during the Great Recession. Importantly, the re-sults of the second-stage regression confirm that establishmentsof more highly levered firms respond more strongly to local con-sumer demand shocks. If anything, the IV estimates are slightlystronger than the OLS estimates. A possible concern with thehousing supply elasticity instrument is that it also includes regu-latory constraints, which may be driven by the same unobservedheterogeneity that also drives employment dynamics. To mitigatethis concern, we repeat the analysis using only the part of the in-strument that is based on an area’s topology, “share of unavailableland.” All results remain similar.

III.D. Industry Sectors

The summary statistics in Table I show that establishmentsof more highly levered firms are underrepresented in the non-tradable sector and overrepresented in the “other” sector—that



is, industries that are neither tradable nor nontradable.9 Whileall our establishment-level regressions include industry fixed ef-fects, we can directly address any concerns related to industrysector composition by performing separate analyses for each in-dustry sector.

Figure II plots the relationship between changes inestablishment-level employment, changes in house prices, andfirm leverage separately for the nontradable and tradable sec-tors.10 The plots for the nontradable sector are similar to those inFigure I. In the lowest leverage quartile, there is a positive albeitweak relationship between changes in house prices and changes inestablishment-level employment (Panel A). In the highest lever-age quartile, this relationship is strongly positive (Panel B). Onthe other hand, there is no clear association between changes inhouse prices and changes in establishment-level employment inthe tradable sector (Panels C and D).

Table IV confirms this visual impression using regres-sion analysis. As is shown, there is a positive and significantcorrelation between changes in house prices and changes inestablishment-level employment in the nontradable sector (col-umn (1)). By contrast, there is no significant correlation in thetradable sector (column (2)). Together, both results confirm sim-ilar results by Mian and Sufi (2014a), who examine changes inaggregate employment at the county level. Although differencesin results across industries are sometimes a concern, the oppositeis true here. Indeed, if changes in house prices affect local em-ployment through changes in consumer demand, then variationin house prices should explain (regional) variation in employmentprimarily in the nontradable sector, where demand by house-holds is local. By contrast, variation in house prices should notcorrelate strongly with variation in employment in the tradable

9. Mian and Sufi (2014a) classify an industry as tradable if imports plus ex-ports exceed $10,000 per worker or $500M in total. Retail industries and restau-rants are classified as nontradable. We label industries that are neither tradablenor nontradable as “other.” The “other” sector comprises a diverse set of industriesthat includes, for example, news and entertainment, transportation and trucking,health care and hospitals, and wholesale. Mian and Sufi also provide an alterna-tive industry classification based on the geographical concentration of industries.Our results are similar when using this alternative classification. See Table 4 ofthe Online Appendix.

10. The plots for the “other” sector are similar to those for the nontradablesector. See Figure 1 of the Online Appendix.




FIG

UR

EII

Non

trad

able

and

Tra

dabl

eE

mpl

oym

ent

atth

eE

stab

lish

men

tL

evel

Th

epl

ots

are

sim

ilar

toth

ose

inP

anel

sA

and

Dof

Fig

ure

I,ex

cept

that

the

sam

ple

isre

stri

cted

ton

ontr

adab

lean

dtr

adab

lein

dust

ries

,res

pect

ivel

y.


TABLE IVTRADABLE AND NONTRADABLE INDUSTRIES

�Log(Emp)07−09

(1) (2) (3) (4)Nontradable Tradable Nontradable Tradable

�Log(HP)06−09 0.074** 0.009 0.029 −0.015(0.035) (0.019) (0.019) (0.043)

�Log(HP)06−09 × 0.131*** 0.037Leverage06 (0.034) (0.120)

Leverage06 −0.038** −0.026(0.015) (0.020)

Industry fixed effects Yes Yes Yes YesR-squared 0.04 0.03 0.04 0.03Observations 124,100 9,900 124,100 9,900

Notes. This table presents variants of the regressions in Table II in which the sample is restricted to tradableand nontradable industries, respectively. Tradable and nontradable industries are described in Mian and Sufi(2014a). All regressions are weighted by establishment size. Standard errors (in parentheses) are clusteredat both the state and firm level. **, and *** denote significance at the 5%, and 1% levels, respectively.

sector, where demand is national or global. Given the evidence inTable IV, as well as further evidence in Mian, Rao, and Sufi (2013),Mian and Sufi (2014a), Stroebel and Vavra (2015), and Kaplan,Mitman, and Violante (2016), we use “falling house prices” and“consumer demand shocks” interchangeably.11

Table 5 of the Online Appendix considers the “other” in-dustry sector. As can be seen, results are similar to those forthe nontradable sector. Indeed, the elasticity of establishment-level employment with respect to house prices is virtually iden-tical in both sectors (0.074 versus 0.075). Together, the nontrad-able and “other” sector account for 97% of all establishment-levelobservations. Hence, there is no need to interact changes in house

11. In principle, falling house prices could affect local employment throughvarious channels. For instance, they could impair the collateral value associatedwith local firms’ commercial real estate or affect local credit supply—for example,local banks reduce lending after experiencing losses on their mortgage loan portfo-lios. Either way, however, this would imply that falling house prices should affectlocal employment also in the tradable sector, contrary to what is observed in thedata. See Mian and Sufi (2014a) for a further dicussion. In addition, Mian, Rao,and Sufi (2013) and Kaplan, Mitman, and Violante (2016) provide direct evidenceshowing that counties or core-based statistical areas (CBSAs) experiencing largerdeclines in housing net worth exhibit larger declines in consumer spending duringthe Great Recession. Likewise, Stroebel and Vavra (2015) find that homeownersbecome more price-sensitive and cut back more on their retail spending in zipcodes experiencing larger drops in house prices.



prices with sector dummies in our regressions.12 Importantly, inboth sectors, the correlation between changes in house prices andchanges in employment is significantly stronger among establish-ments of more highly levered firms. Accordingly, our results arenot driven by industry sector composition effects.

Table 7 of the Online Appendix lists the top 10 industriesin which house prices have the biggest impact on establishment-level employment. To construct this list, we estimated column (1)of Table II separately for each four-digit NAICS industry. At thetop of the list are full-service restaurants (nontradable), buildingmaterial and supplies dealers (“other”), and health and personalcare stores (nontradable). Interestingly, 3 of the top 10 industriesare auto-related: automotive repair and maintenance (#4, “other”);automotive parts, accessories, and tire stores (#7, nontradable);and automobile dealers (#8, nontradable). Not surprisingly, thereis no tradable industry in the top 10.13

III.E. Establishment Closures

Does the drop in house prices between 2006 and 2009 causefirms to make adjustments at the extensive margin? To exam-ine this question, we include in our sample establishments thatare closed down between 2007 and 2009. The dependent variableis a dummy indicating whether an establishment is closed dur-ing that period. As is shown in Table 8 of the Online Appendix,changes in house prices between 2006 and 2009 are negativelyand significantly associated with establishment closures (column(1)). Moreover, as in our employment regressions, this effect is sig-nificantly stronger among establishments of more highly leveredfirms (column (2)). Hence, firms respond to falling house prices bymaking adjustments at both the intensive and extensive margin.

12. While tradable industries account for 3% of all establishments, they ac-count for 12% of total employment in our sample. In other words, firms in tradableindustries have relatively few but large establishments (for example, manufactur-ing plants). Since all our regressions are employment-weighted, this implies thatexcluding tradable industries should make our results stronger. Indeed, Table 6 ofthe Online Appendix shows that our results become slightly stronger if we excludetradable industries.

13. Cement and concrete product manufacturing (#10, “other”) is not classifiedas tradable because its imports plus exports do not exceed $10,000 per worker or$500M in total. Due to excessively high transportation costs, the market for cementand concrete manufacturing is largely local.





III.F. Compustat-LBD Sample versus Full LBD Sample

Our sample consists of establishments in the LBD—for ex-ample, retail stores, supermarkets, or restaurants—whose par-ent firms have a match in Compustat. Thus, our sample doesnot include establishments of private firms or, more important,single-unit establishments (e.g., “mom and pop shops”).14 Overall,our sample accounts for 12% of total LBD employment. In termsof industry sectors, our sample accounts for 26% of nontradableemployment, 18% of tradable employment, and 8% of “other” em-ployment.

One might worry that our sample consists of establishmentsthat are particularly responsive to local consumer demand shocks.In Table 9 of the Online Appendix, we estimate the elasticity ofestablishment-level employment with respect to house prices sep-arately for establishments in the matched Compustat-LBD sam-ple and those in the full LBD sample. As is shown, the elasticity is59% larger in the full LBD sample.15 Thus, if anything, our sam-ple includes establishments that respond less strongly to localconsumer demand shocks.

That establishments in the matched Compustat-LBD sam-ple have lower elasticities is consistent with Compustat firms be-ing less financially constrained. Indeed, several empirical studiesprovide evidence suggesting that public firms are less financiallyconstrained than private firms, for example, Brav (2009) and Giljeand Taillard (2015). Notably, the lower elasticities are not due toCompustat firms being located in regions with smaller house pricedrops: the correlation between �Log(HP)06−09 and the employ-ment share of Compustat firms at either the zip code or countylevel is close to zero and insignificant (1.4% [p-value: .321] and1.2% [p-value: .681]). Likewise, a Kolmogorov-Smirnov test is un-able to reject the null that the distribution of �Log(HP)06−09 isidentical for establishments in the matched Compustat-LBD sam-ple and other establishments in the LBD. Thus, establishmentsin the matched Compustat-LBD sample and other establishmentsin the LBD experience similar declines in house prices.

14. Our county-level analysis in Section VI constitutes an exception. There,some of our regressions have total LBD employment as the dependent variable.

15. A main difference between the two samples is firm size: Compustat firmsare much larger. Indeed, if we reweight our regressions using firm size instead ofestablishment size—thus giving more weight to establishments of larger firms—the elasticity in the full LBD sample is only 24% larger than in the matchedCompustat-LBD sample. See Panel C of Table 9 in the Online Appendix.




IV. FINANCIAL CONSTRAINTS AND LABOR HOARDING

The concept of labor hoarding, which goes back to the early1960s, posits that firms facing a temporary (e.g., cyclical) drop indemand choose to retain more workers than would be technicallynecessary so as to economize on the costs of firing, hiring, andtraining workers. Direct evidence in support of labor hoardingcomes from a survey of plant managers by Fay and Medoff (1985)asking detailed questions about the workforce retained during theplant’s most recent downturn. The typical plant paid for about 8%more blue-collar labor hours in a downturn than were techni-cally necessary to meet production requirements. About half ofthis labor could be justified by other useful tasks—for example,maintenance, cleaning, or training—leaving about 4% of the blue-collar hours paid for by the typical plant to be classified as truly“hoarded.” By the 1980s and 1990s, the concept of labor hoardinghad become an “accepted part of economists’ explanations of theworkings of labor markets and of the relationship between laborproductivity and economic fluctuations” (Biddle 2014, p. 197).16

Labor hoarding is costly, however. Effectively, firms must(temporarily) subsidize workers’ wages. Hence, firms with lit-tle financial slack face a genuine trade-off between long-runoptimization—saving on the costs of firing, hiring, and (re-)training workers—and short-run liquidity needs. Our results sug-gest that firms with weak balance sheets—and tighter financialconstraints—are more apt to respond to this trade-off by engagingin less labor hoarding. In other words, firms with weak balancesheets cut more jobs in response to a decline in consumer demandthan they (optimally) would have in the absence of financial con-straints.

In our sample, more highly levered firms indeed appear to bemore financially constrained. According to the summary statis-tics in Table I, they score worse on popular measures of finan-cial constraints, such as the indexes by Kaplan and Zingales(1997) and Whited and Wu (2006). But do they also act like fi-nancially constrained firms during the Great Recession? To ad-dress this question, we turn to firm-level regressions. Precisely,we estimate the firm-level analogue of our baseline specification,where �Log(HP)06−09 is now the employment-weighted average

16. Biddle (2014) provides a comprehensive overview of the labor hoardingliterature.


percentage change in house prices between 2006 and 2009 acrossall of the firm’s establishments. In other words, �Log(HP)06−09is the average consumer demand shock faced by the firm. Thedependent variable at the firm level is either the change in short-term debt, long-term debt, or equity, the change in employment orinvestment, or the fraction of establishments closed, all between2007 and 2009. The first three dependent variables measure afirm’s access to external finance during the Great Recession. Thelast three dependent variables measure if being financially con-strained has real consequences at the firm level.

Table V presents the results. When faced with consumer de-mand shocks in the Great Recession, more highly levered firmsare less apt (or able) to raise additional short- and long-term debt(columns (1) and (2)).17 As a consequence, they experience morelayoffs, are more likely to close down establishments, and cut backmore on investment (columns (4) to (6)). Overall, these results sug-gest that firms with higher leverage not only appear to be morefinancially constrained but also act like financially constrainedfirms during the Great Recession.

We should note that ours is not the first publication to point toa link between financial constraints and labor hoarding.18 Usingmanufacturing firm-level data from 1959 to 1985, Sharpe (1994)found that employment growth is more cyclical at more highlylevered firms. As we do, Sharpe concludes that financial con-straints impair firms’ ability to engage in labor hoarding. Surveyevidence by Campello, Graham, and Harvey (2010) supports thisconclusion. The authors asked 574 U.S. CFOs in 2008 whethertheir firms are financially constrained and what they are plan-ning to do in 2009. Firms classified as financially constrained saidthey would cut their employment by 10.9% in the following year.By contrast, firms classified as unconstrained said they wouldcut their employment only by 2.7%. Although both studies sug-gest a link between employment growth and financial constraints

17. We are unable to reject the null that �Log(HP)06−09 + �Log(HP)06−09× Leverage06 evaluated at Leverage06 = 1 is 0 in columns (1) and (2) (p-values:.333 and .268, respectively). Accordingly, firms with very high leverage do not,or cannot, raise any short- or long-term debt when faced with consumer demandshocks in the Great Recession.

18. We are unaware of theory models linking financial constraints and laborhoarding. For theory models of labor hoarding per se—that is, absent financialfrictions—see Oi (1962), Clark (1973), Rotemberg and Summers (1990), and Burn-side, Eichenbaum, and Rebelo (1993).


TA

BL

EV

FIR

M-L

EV

EL

AN

ALY

SIS

Ext

ern

alfi

nan

ceE

mpl

oym

ent

and

inve

stm

ent

�S

Tde

bt07

−09

�LT

debt

07−0

9�

Equ

ity 0

7−09

�L

og(E

mp)

07−0

9E

st.C

losu

re07

−09

�C

AP

EX

07−0

9(1

)(2

)(3

)(4

)(5

)(6

)

�L

og(H

P) 0

6−09

−0.0

25**

−0.0

40**

0.00

50.

020

−0.0

080.

002

(0.0

11)

(0.0

19)

(0.0

37)

(0.0

33)

(0.0

15)

(0.0

05)

�L

og(H

P) 0

6−09

×L

ever

age 0

60.

035*

*0.

059*

*−0

.011

0.12

2***

−0.0

46**

0.01

4**

(0.0

14)

(0.0

21)

(0.0

47)

(0.0

40)

(0.0

19)

(0.0

07)

Lev

erag

e 06

−0.0

11*

−0.0

19*

0.00

9−0

.024

**0.

018*

*−0

.005

*(0

.006

)(0

.011

)(0

.021

)(0

.010

)(0

.008

)(0

.003

)In

dust

ryfi

xed

effe

cts

Yes

Yes

Yes

Yes

Yes

Yes

R-s

quar

ed0.

090.

080.

030.

110.

110.

14O

bser

vati

ons

2,80

02,

800

2,80

02,

800

2,80

02,

800

Not

es.T

his

tabl

epr

esen

tsfi

rm-l

evel

vari

ants

ofth

ere

gres

sion

sin

Tab

leII

.Sh

ort-

term

(ST

)de

btis

the

rati

oof

debt

incu

rren

tli

abil

itie

sdi

vide

dby

tota

las

sets

.Lon

g-te

rm(L

T)

debt

isth

era

tio

oflo

ng-

term

debt

divi

ded

byto

tal

asse

ts.

Equ

ity

isth

era

tio

ofth

ebo

okva

lue

ofeq

uit

ydi

vide

dby

tota

las

sets

.E

stab

lish

men

t(E

st.)

clos

ure

isth

en

um

ber

ofes

tabl

ish

men

tscl

osed

betw

een

2007

and

2009

divi

ded

byth

en

um

ber

ofes

tabl

ish

men

tsin

2007

.CA

PE

Xis

the

rati

oof

capi

tale

xpen

ditu

res

topr

oper

ty,p

lan

tan

deq

uip

men

t(P

P&

E).

�L

og(H

P) 0

6−09

isag

greg

ated

atth

efi

rmle

velb

yco

mpu

tin

gth

eem

ploy

men

t-w

eigh

ted

aver

age

valu

eof

�L

og(H

P) 0

6−09

acro

ssal

loft

he

firm

’ses

tabl

ish

men

ts.I

ndu

stry

fixe

def

fect

sar

eba

sed

onfo

ur-

digi

tN

AIC

Sco

des.

Sta

nda

rder

rors

(in

pare

nth

eses

)ar

ecl

ust

ered

atth

est

ate

leve

l.*,

**,a

nd

***

den

ote

sign

ifica

nce

atth

e10

%,5

%,a

nd

1%le

vels

,res

pect

ivel

y.


over the business cycle, neither study separates out the effects ofdemand shocks that lie at the very heart of the labor hoardingconcept.

V. ROBUSTNESS

An important concern is that more highly levered firms re-spond more strongly to local consumer demand shocks not becausethey are more financially constrained but rather because of otherreasons unrelated to financial constraints. Although we cannotrule out this possibility in general, we can address specific alter-native stories.19 For instance, according to the summary statisticsin Table I, more highly levered firms are less productive and ex-panded more in the years before the Great Recession. However,this raises the concern that these firms respond more strongly toconsumer demand shocks during the Great Recession not becausethey are more financially constrained but because they are lessproductive or expanded too much in previous years. We examinethese and other alternative stories in Table VI by including addi-tional controls Z and �Log(HP)06−09 × Z in our regressions, whereZ stands for the alternative hypothesis in question. For brevity,Table VI only shows the coefficients associated with the main vari-ables of interest, �Log(HP)06−09 × Leverage06 and �Log(HP)06−09× Z. The full sets of coefficients are provided in Tables 10 to 14 ofthe Online Appendix.

V.A. Employment and Asset Growth

Panel A examines if our results are driven by firms expand-ing too much in the years prior to the Great Recession. In column(1), we include as additional controls the percentage change inemployment between 2002 and 2006, �Log(Emp)02−06, as wellas its interaction with �Log(HP)06−09. Column (2) is similar, ex-cept that we consider the percentage change in assets between2002 and 2006, �Log(Assets)02−06. As it turns out, including past

19. Some alternative stories can be ruled out a priori. For instance,Table I shows that there is virtually no correlation between firm leverage andeither housing supply elasticity or changes in house prices during the Great Reces-sion. Hence, more highly levered firms do not respond more strongly to consumerdemand shocks because they are located in regions with stronger shocks. Indeed,given our fixed-effect specification in Table II, we can rule out any alternative storywhereby low- and high-leverage firms differ along either geographical or industrydimensions.


FIRM LEVERAGE, CONSUMER DEMAND, EMPLOYMENT LOSSES297T

AB

LE

VI

RO

BU

ST

NE

SS

(1)

(2)

(3)

(4)

Pan

elA

:Em

ploy

men

tan

das

set

grow

th�

Log

(HP

) 06−

09×

Lev

erag

e 06

0.11

1**

0.11

3***

(0.0

47)

(0.0

40)

�L

og(H

P) 0

6−09

×�

Log

(Em

p)02

−06

0.02

7(0

.034

)�

Log

(HP

) 06−

09×

�L

og(A

sset

s)02

−06

0.01

2(0

.009

)P

anel

B:P

rodu

ctiv

ity

�L

og(H

P) 0

6−09

×L

ever

age 0

60.

101*

*0.

122*

**0.

108*

**(0

.040

)(0

.041

)(0

.039

)�

Log

(HP

) 06−

09×

RO

A06

−0.0

24(0

.110

)�

Log

(HP

) 06−

09×

NP

M06

−0.0

38(0

.161

)�

Log

(HP

) 06−

09×

TF

P06

−0.0

49*

(0.0

26)

Pan

elC

:Sen

siti

vity

toag

greg

ate

empl

oym

ent

and

hou

sepr

ices

�L

og(H

P) 0

6−09

×L

ever

age 0

60.

109*

*0.

110*

*0.

119*

**0.

120*

**(0

.044

)(0

.044

)(0

.044

)(0

.044

)�

Log

(HP

) 06−

09×

Ela

stic

ity E

mp,

10−y

ear

0.00

6*(0

.004

)�

Log

(HP

) 06−

09×

Ela

stic

ity E

mp,

20−y

ear

0.00

5(0

.004

)�

Log

(HP

) 06−

09×

Ela

stic

ity H

P,10

−yea

r0.

006

(0.0

08)

�L

og(H

P) 0

6−09

×E

last

icit

y HP,

02−0

60.

007

(0.0

08)


TA

BL

EV

I( C

ON

TIN

UE

D)

(1)

(2)

(3)

(4)

Pan

elD

:Act

ivis

tin

vest

ors

�L

og(H

P) 0

6−09

×L

ever

age 0

60.

110*

**0.

115*

**(0

.039

)(0

.040

)�

Log

(HP

) 06−

09×

Hed

geF

un

d 06

0.03

6(0

.036

)�

Log

(HP

) 06−

09×

Pri

vate

Equ

ity 0

60.

009

(0.0

56)

Not

es.

Th

ista

ble

pres

ents

vari

ants

ofth

ere

gres

sion

sin

Tab

leII

inw

hic

hZ

and

�L

og(H

P)0

6–09

×Z

are

incl

ude

das

addi

tion

alco

ntr

ols.

For

brev

ity,

the

tabl

eon

lysh

ows

the

coef

fici

ents

asso

ciat

edw

ith

the

mai

nva

riab

les

ofin

tere

st,�

Log

(HP

)06–

09×

Lev

erag

e06

and

�L

og(H

P)0

6–09

×Z

.T

he

full

sets

ofco

effi

cien

tsar

epr

ovid

edin

Tab

les

10to

14of

the

On

lin

eA

ppen

dix.

InP

anel

A,Z

isei

ther

the

grow

thin

firm

-lev

elem

ploy

men

tfr

om20

02to

2006

,�L

og(E

mp)

02−0

6(c

olu

mn

(1))

,or

the

grow

thin

firm

-lev

elas

sets

from

2002

to20

06,�

Log

(Ass

ets)

02−0

6(c

olu

mn

(2))

.In

Pan

elB

,Zis

eith

erth

efi

rm’s

retu

rnon

asse

ts,R

OA

06(c

olu

mn

(1))

,th

efi

rm’s

net

profi

tm

argi

n,N

PM

06(c

olu

mn

(2))

,or

the

firm

’sto

tal

fact

orpr

odu

ctiv

ity,

TF

P06

(col

um

n(3

)),a

llm

easu

red

in20

06.R

OA

06is

the

rati

oof

oper

atin

gin

com

ebe

fore

depr

ecia

tion

toto

tala

sset

s,N

PM

06is

the

rati

oof

oper

atin

gin

com

ebe

fore

depr

ecia

tion

tosa

les,

and

TF

P06

isth

ere

sidu

alfr

oma

regr

essi

onof

Log

(Sal

es)

onL

og(E

mpl

oyee

s)an

dL

og(P

P&

E)

acro

ssal

lC

ompu

stat

firm

sin

the

sam

etw

o-di

git

SIC

indu

stry

.In

Pan

elC

,Z

isei

ther

the

10-y

ear

elas

tici

tyof

firm

-lev

elem

ploy

men

tw

ith

resp

ect

toag

greg

ate

empl

oym

ent,

Ela

stic

ity E

mp,

10-y

ear(

colu

mn

(1))

,th

e20

-yea

rel

asti

city

offi

rm-l

evel

empl

oym

ent

wit

hre

spec

tto

aggr

egat

eem

ploy

men

t,E

last

icit

y Em

p,20

-yea

r(c

olu

mn

(2))

,th

e10

-yea

rel

asti

city

offi

rm-l

evel

empl

oym

ent

wit

hre

spec

tto

hou

sepr

ices

,Ela

stic

ity H

P,1

0-ye

ar(c

olu

mn

(3))

,or

the

elas

tici

tyof

firm

-lev

elem

ploy

men

tw

ith

resp

ect

toh

ouse

pric

esdu

rin

gth

e20

02to

2006

hou

sin

gbo

om,E

last

icit

y HP

,02−

06(c

olu

mn

(4))

.Ela

stic

ity E

mp,

10-y

ear

and

Ela

stic

ity E

mp,

20-y

ear

are

com

pute

dby

esti

mat

ing

afi

rm-l

evel

regr

essi

onof

�L

og(E

mpl

oym

ent)

ona

con

stan

tan

d�

Log

(LB

DE

mpl

oym

ent)

usi

ng

all

avai

labl

eye

ars

from

1997

to20

06an

d19

87to

2006

,res

pect

ivel

y.E

last

icit

y HP

,10-

year

and

Ela

stic

ity H

P,0

2−06

are

com

pute

das

the

empl

oym

ent-

wei

ghte

dav

erag

eel

asti

city

ofem

ploy

men

tw

ith

resp

ect

toh

ouse

pric

esac

ross

allo

fth

efi

rm’s

esta

blis

hm

ents

,wh

ere

the

latt

eris

com

pute

dei

ther

byes

tim

atin

gan

esta

blis

hm

ent-

leve

lreg

ress

ion

of�

Log

(Em

ploy

men

t)on

aco

nst

ant

and

�L

og(H

P)

usi

ng

all

avai

labl

eye

ars

from

1997

to20

06(E

last

icit

y HP

,10-

year

)or

asth

epe

rcen

tage

chan

gein

esta

blis

hm

ent-

leve

lem

ploy

men

tdi

vide

dby

the

perc

enta

gech

ange

inh

ouse

pric

esdu

rin

gth

e20

02to

2006

hou

sin

gbo

om(E

last

icit

y HP

,02−

06).

InP

anel

D,Z

isei

ther

adu

mm

yin

dica

tin

gw

het

her

afi

rmis

targ

eted

byan

acti

vist

hed

gefu

nd,

Hed

geF

un

d 06

(col

um

n(1

)),o

ra

dum

my

indi

cati

ng

wh

eth

era

firm

has

sign

ifica

nt

(i.e

.,m

ore

than

5%)

priv

ate

equ

ity

own

ersh

ip,P

riva

teE

quit

y 06

(col

um

n(2

)),a

llm

easu

red

in20

06.I

ndu

stry

fixe

def

fect

sar

eba

sed

onfo

ur-

digi

tN

AIC

Sco

des.

All

regr

essi

ons

are

wei

ghte

dby

esta

blis

hm

ent

size

.Sta

nda

rder

rors

(in

pare

nth

eses

)ar

ecl

ust

ered

atbo

thth

est

ate

and

firm

leve

l.*,

**,a

nd

***

den

ote

sign

ifica

nce

atth

e10

%,5

%,a

nd

1%le

vels

,res

pect

ivel

y.



employment or asset growth has no effect on our results. Al-though none of the additional controls are significant, the maincoefficient of interest—that associated with the interaction term�Log(HP)06−09 × Leverage06—is significant and of similar mag-nitude as in our main specification in Table II.

V.B. Productivity

Panel B examines if our results are driven by firms havinglower productivity. In column (1), we include as additional con-trols the firm’s return on assets, ROA06, as well as its interactionwith �Log(HP)06−09. Columns (2) and (3) are similar, except thatwe consider the firm’s net profit margin, NPM06, and total factorproductivity, TFP06, respectively. All three productivity measuresare measured in 2006. In all three cases, the main effect of pro-ductivity is significant with the predicted sign: less productivefirms experience larger job losses during the Great Recession.However, the interaction with �Log(HP)06−09, while having thepredicted sign, is either insignificant (columns (1) and (2)) or onlyweakly significant (column (3)). Importantly, the coefficient asso-ciated with the interaction term �Log(HP)06−09 × Leverage06 issignificant and stable in all three regressions.

V.C. Sensitivity to Aggregate Employment and House Prices

Panel C examines if our results are driven by firms being gen-erally more sensitive to either aggregate employment or houseprices (“high-beta firms”)—that is, for reasons unrelated to fi-nancial constraints.20 We can separately identify both effects be-cause firms with weak balance sheets in the Great Recession maynot have been firms with weak balance sheets in prior down-turns. In columns (1) and (2), we include as additional controlsthe elasticity of firm-level employment with respect to total LBDemployment over either a 10-year or 20-year period ending in2006, ElasticityEmp, 10-year and ElasticityEmp, 20-year, respectively, aswell as its interaction with �Log(HP)06−09. Columns (3) and (4)are similar, except that we consider the elasticity of firm-levelemployment with respect to house prices over either a 10-year

20. For example, high-leverage firms may have customers that are more aptto switch to alternative (e.g., cheaper) brands during economic downturns. In thatcase, firms with higher leverage may have a higher sensitivity to demand shocks,albeit for reasons unrelated to financial constraints.


period ending in 2006 or during the 2002 to 2006 housing boom,ElasticityHP,10-year and ElasticityHP,02-06, respectively.21

As is shown, the main effect of the elasticity of firm-level em-ployment with respect to either aggregate employment or houseprices is always significant with the predicted sign: firms that hadbeen previously more sensitive to either aggregate employment orhouse prices also experience greater employment losses during theGreat Recession. However, the interaction with � Log(HP)06−09,while having the predicted sign, is either only weakly significant(column (1)) or insignificant (columns (2) to (4)). Importantly, ourmain coefficient of interest—that associated with the interactionterm �Log(HP)06−09 × Leverage06—is significant and stable in allfour regressions.

V.D. Activist Investors

Panel D examines if our results are driven by firms being tar-geted by activist investors, such as activist hedge funds or privateequity (PE) funds. The hedge fund data come from an extendedversion of the data set used in Brav et al. (2008). The data arebased on Schedule 13D filings, which investors must file withthe SEC within 10 days of acquiring more than 5% of any classof securities of a publicly traded company if they have an inter-est in influencing the management of the company. The PE dataare obtained from Thomson Reuter’s 13F database, which reportsquarterly holdings for all institutional owners with an ownershipstake of at least 5%. The names of the owners are matched tothe list of PE firms obtained from VentureXpert using a fuzzymatching algorithm. All matches are reviewed by hand to ensureaccuracy.

In column (1), we include as additional controls a dummyvariable indicating whether a firm is targeted by an activisthedge fund in 2006, Hedge Fund06, as well as its interactionwith �Log(HP)06−09. Column (2) is similar, except that we includea dummy variable indicating whether a firm has significant PEownership in 2006, PE06.22 As is shown, the main effect of activist

21. Table 13 of the Online Appendix additionally considers the elasticity offirm-level employment with respect to aggregate employment over either a 15-year or 30-year period as well as the elasticity of firm-level employment withrespect to either aggregate employment or house prices during the 2001 recession.

22. We obtain similar results if we include dummies indicating involvementby activist investors over the 2002 to 2006 period, Hedge Fund02−06 and PE02−06,respectively.



investors has the predicted sign—firms with activist investors cutmore jobs during the Great Recession—but is either only weaklysignificant (column (1)) or insignificant (column (2)). The inter-action with �Log(HP)06−09, while having the predicted sign, isalways insignificant. Importantly, the coefficient associated withthe interaction term �Log(HP)06−09 × Leverage06 is significantand stable in both regressions.

VI. COUNTY-LEVEL ANALYSIS

VI.A. General Equilibrium Adjustments

In general equilibrium, output and labor may shift from high-to low-leverage firms. The magnitude of this reallocation dependson how much output prices at low-leverage firms decline relativeto those at high-leverage firms as well as the substitutability ofgoods between the firms. If prices are sticky or goods are imperfectsubstitutes, the magnitude of this reallocation is limited, with theimplication that the distribution of firm leverage matters also inthe aggregate. If there is trade across regions, output and labormay not only shift to low-leverage firms but more generally tofirms that are less prone to local demand shocks, such as firmsin the tradable sector. Again, the magnitude of this reallocationwill depend on how sticky prices are. Along those lines, Gilchristet al. (2016), Mian and Sufi (2014a), and Chodorow-Reich (2014)all discuss how sticky prices or imperfect substitutability in thegoods market may limit the reallocation of output and labor inresponse to differential firm-level shocks.23

The extent of labor reallocation also depends on search andmatching frictions and labor adjustment costs. Some evidence sug-gests that labor market frictions were particularly severe dur-ing the Great Recession. Davis (2011) and Davis, Faberman, andHaltiwanger (2013) find that both the search intensity per unem-ployed worker and the recruiting intensity per vacancy droppedsharply during the Great Recession. Likewise, Sahin et al. (2014)find that mismatch between job seekers and vacant jobs increasedmarkedly during the Great Recession, explaining up to one third ofthe rise in unemployment. Overall, Foster, Grim, and Haltiwanger

23. Besides, if low-leverage firms engage in labor hoarding (see Section IV),this means their own workforce is employed at below full capacity, making it lesslikely that these firms would want to hire additional workers from high-leveragefirms.


(2016) find that the intensity of labor reallocation fell rather thanrose during the Great Recession, contrary to prior recessions. Theyconclude that “job reallocation (creation plus destruction) is at itslowest point in 30 years during the Great Recession and its imme-diate aftermath” (Foster, Grim, and Haltiwanger 2016, p. S305).

VI.B. Firm Leverage, House Prices, and Employment at theCounty Level

To empirically investigate whether the distribution of firmleverage matters in the aggregate, we turn to county-level re-gressions. Imagine two counties, one with a smaller (employment-weighted) share of high-leverage firms and the other with a largershare. Suppose further that both counties exhibit a similar dropin house prices. If our previous results also hold in the aggre-gate, then the more highly levered county should experience alarger decline in county-level employment. By contrast, if the dis-tribution of firm leverage does not matter in the aggregate, thenboth counties should experience similar declines in county-levelemployment, regardless of the level of county leverage.

We proceed as in our establishment-level analysis. County-level employment is total employment by all firms in oursample within a county. County-level leverage is the employment-weighted average value of Leverage06 across all firms in our sam-ple within a county. Importantly, county-level leverage is uncor-related with changes in county-level house prices between 2006and 2009, �Log(HP)06−09: the correlation is 0.8% and highly in-significant (p-value: .809). Thus, low- and high-leverage countiesexperience similar consumer demand shocks during the Great Re-cession.

To obtain a visual impression, Figure III plots the percent-age change in county-level employment between 2007 and 2009,�Log(Emp)07−09, against the percentage change in county-levelhouse prices between 2006 and 2009, �Log(HP)06−09, separatelyfor counties in the lowest and highest leverage quartile. For eachpercentile of �Log(HP)06−09, the scatterplot shows the mean val-ues of �Log(HP)06−09 and �Log(Emp)07−09, respectively. The plotsin Panels A and B (all industry sectors) are similar to those at theestablishment level (see Figure I, Panels A and D). In Panel A,which represents the lowest leverage quartile, there is a posi-tive but weak relationship between changes in house prices andchanges in county-level employment. In Panel B, which represents


FIG

UR

EII

I

Fir

mL

ever

age,

Hou

seP

rice

s,an

dE

mpl

oym

ent

atth

eC

oun

tyL

evel

Th

epl

ots

are

sim

ilar

toth

ose

inP

anel

sA

and

Dof

Fig

ure

Ian

dP

anel

sA

toD

inF

igu

reII

,ex

cept

that

the

anal

ysis

isat

the

cou

nty

leve

l.C

oun

ty-l

evel

empl

oym

ent

isto

tal

empl

oym

ent

byal

lfi

rms

inou

rsa

mpl

ew

ith

ina

cou

nty

.C

oun

ty-l

evel

leve

rage

isth

eem

ploy

men

t-w

eigh

ted

aver

age

valu

eof

firm

leve

rage

acro

ssal

lfirm

sin

our

sam

ple

wit

hin

aco

un

ty.


FIG

UR

EII

I

(CO

NT

INU

ED

)


the highest leverage quartile, this relationship becomes stronglypositive. The elasticity of county-level employment with respectto house prices in the highest leverage quartile is 0.095, whichis more than seven times larger than the elasticity in the lowestleverage quartile and virtually identical to the corresponding elas-ticity of 0.096 at the establishment level (see Figure I, Panel D).The plots in Panels C to F are also similar to those at the es-tablishment level (see Figure II, Panels A–D). In Panels C and D(nontradable sector), there is a positive but weak relationshipbetween changes in house prices and changes in county-level em-ployment in lowest leverage quartile and a strongly positive rela-tionship in the highest leverage quartile. By contrast, in PanelsE and F (tradable sector), there is no clear association betweenchanges in house prices and changes in county-level employment.

Table VII confirms this visual impression using regres-sion analysis. To facilitate comparison with our establishment-level regressions, we include as controls the county-level em-ployment shares of all two-digit NAICS industries in 2006 (seeTable III of Mian and Sufi 2014a). Columns (1) and (2) ex-amine county-level employment across all industry sectors. Incolumn (1), the average elasticity of county-level employmentwith respect to house prices is 0.069, which is similar to theelasticity of 0.066 in our establishment-level analysis (see Ta-ble II, column (1)).24 Importantly, in column (2), the interactionterm �Log(HP)06−09× Leverage06 is highly significant, and itscoefficient of 0.110 is similar to the coefficient of 0.114 at the estab-lishment level (see Table II, column (3)). Thus, more highly leveredcounties exhibit significantly larger declines in employment in re-sponse to consumer demand shocks, implying that the distributionof firm leverage matters also in the aggregate. Columns (3) and(4) examine county-level employment in the nontradable sector.In column (3), the average elasticity of nontradable county-levelemployment with respect to house prices is 0.076, which is sim-ilar to the elasticity of 0.074 in our establishment-level analysis(see Table IV, column (1)). Moreover, in column (4), the interac-tion term �Log(HP)06−09 × Leverage06 is highly significant, andits coefficient of 0.134 is similar to the coefficient of 0.131 at the

24. The elasticity of 0.066 in column (1) of Table II is estimated withoutindustry fixed effects. The corresponding elasticity with industry fixed effects is0.068, which is almost identical to the county-level elasticity of 0.069 (see Table 9,Panel A, column (1) of the Online Appendix).



TA

BL

EV

IIC

OU

NT

Y-L

EV

EL

AN

ALY

SIS �

Log

(Em

p)07

−09

All

indu

stri

esN

ontr

adab

leT

rada

ble

(1)

(2)

(3)

(4)

(5)

(6)

�L

og(H

P) 0

6−09

0.06

9***

0.02

50.

076*

**0.

026

0.01

0−0

.015

(0.0

08)

(0.0

19)

(0.0

12)

(0.0

20)

(0.0

10)

(0.0

39)

�L

og(H

P) 0

6−09

×L

ever

age 0

60.

110*

**0.

134*

**0.

034

(0.0

38)

(0.0

36)

(0.1

51)

Lev

erag

e 06

−0.0

29**

−0.0

39**

−0.0

28*

(0.0

13)

(0.0

20)

(0.0

15)

Cou

nty

×in

dust

ryco

ntr

ols

Yes

Yes

Yes

Yes

Yes

Yes

R-s

quar

ed0.

130.

130.

180.

190.

090.

09O

bser

vati

ons

1,00

01,

000

1,00

01,

000

1,00

01,

000

Not

es.T

his

tabl

epr

esen

tsco

un

ty-l

evel

vari

ants

ofth

ere

gres

sion

sin

Tab

leII

and

IV.C

oun

ty-l

evel

empl

oym

ent

isto

tal

empl

oym

ent

byal

lfi

rms

inou

rsa

mpl

ew

ith

ina

cou

nty

.C

oun

ty-l

evel

leve

rage

isth

eem

ploy

men

t-w

eigh

ted

aver

age

valu

eof

Lev

erag

e 06

acro

ssal

lfi

rms

inou

rsa

mpl

ew

ith

ina

cou

nty

.Cou

nty

×in

dust

ryco

ntr

ols

are

the

cou

nty

-spe

cifi

cem

ploy

men

tsh

ares

ofal

l23

two-

digi

tN

AIC

Sin

dust

ries

in20

06.A

llre

gres

sion

sar

ew

eigh

ted

byco

un

tysi

ze.S

tan

dard

erro

rs(i

npa

ren

thes

es)

are

clu

ster

edat

the

stat

ele

vel.

*,**

,an

d**

*de

not

esi

gnifi

can

ceat

the

10%

,5%

,an

d1%

leve

ls,r

espe

ctiv

ely.


establishment level (see Table IV, column (3)). Finally, columns (5)and (6) examine county-level employment in the tradable sector.Regardless of county-level leverage, there is no significant associ-ation between changes in house prices and changes in county-levelemployment, confirming similar results in our establishment-level analysis (see Table IV, columns (2) and (4)).

Tables 15 to 19 of the Online Appendix provide the samerobustness tests as in Section V, except that the analysis is atthe county level. There is no evidence that more highly leveredcounties respond more strongly to consumer demand shocks dur-ing the Great Recession because their firms are less productive,have expanded too much prior to the Great Recession, or havemore activist investors, or because employment in these countiesis generally more sensitive to either aggregate employment orhouse prices.

VI.C. Expanding the Sample: County-Level Employmentby All Firms in the LBD

Examining changes in county-level employment by all firmsin our sample captures any general equilibrium effects from laborreallocation between firms in our sample within a given county.However, it does not capture labor reallocation between firms inour sample and other firms in the LBD, such as small private firmsand mom and pop shops. Given that the latter firms exhibit evenlarger employment losses in response to consumer demand shocks(see Table 9 of the Online Appendix), one would think that laborreallocation toward these firms is unlikely. Ultimately, however,this is an empirical question. In the remainder of this section, wethus examine changes in county-level employment by all firms inthe LBD.

Expanding the sample to include all firms in the LBD entailstwo main changes. The first change is that firms become more sen-sitive to changes in house prices. As discussed in Section III.F, theaverage firm in the LBD is more financially constrained than theaverage firm in the matched Compustat-LBD sample. Accordingly,we would expect that the average elasticity of county-level employ-ment with respect to house prices increases. Indeed, as column (1)of Table VIII shows, the average elasticity of county-level employ-ment with respect to house prices is now 0.104, which is about50% larger than the average elasticity of 0.069 in the matchedCompustat-LBD sample (see column (1) of Table VII). Note that




TA

BL

EV

III

EX

PA

ND

ING

TH

ES

AM

PL

E:C

OU

NT

Y-L

EV

EL

EM

PL

OY

ME

NT

BY

AL

LF

IRM

SIN

TH

EL

BD

�L

og(E

mp)

07−0

9

All

indu

stri

esN

ontr

adab

leT

rada

ble

(1)

(2)

(3)

(4)

(5)

(6)

�L

og(H

P) 0

6−09

0.10

4***

0.08

5***

0.12

2***

0.09

6***

0.00

8−0

.003

(0.0

05)

(0.0

15)

(0.0

06)

(0.0

19)

(0.0

16)

(0.0

34)

�L

og(H

P) 0

6−09

×L

ever

age 0

60.

045*

*0.

058*

**0.

022

(0.0

22)

(0.0

20)

(0.1

05)

Lev

erag

e 06

−0.0

23**

−0.0

28**

−0.0

19*

(0.0

10)

(0.0

14)

(0.0

10)

Cou

nty

×in

dust

ryco

ntr

ols

Yes

Yes

Yes

Yes

Yes

Yes

R-s

quar

ed0.

140.

140.

160.

160.

130.

13O

bser

vati

ons

1,00

01,

000

1,00

01,

000

1,00

01,

000

Not

es.T

his

tabl

epr

esen

tsva

rian

tsof

the

regr

essi

ons

inT

able

VII

inw

hic

hco

un

ty-l

evel

empl

oym

ent

isto

tal

empl

oym

ent

byal

lfi

rms

inth

eL

BD

wit

hin

aco

un

ty.C

oun

ty-l

evel

leve

rage

isth

eem

ploy

men

t-w

eigh

ted

aver

age

valu

eof

Lev

erag

e 06ac

ross

all

firm

sin

our

sam

ple

wit

hin

aco

un

ty.

Cou

nty

×in

dust

ryco

ntr

ols

are

the

cou

nty

-spe

cifi

cem

ploy

men

tsh

ares

ofal

l23

two-

digi

tN

AIC

Sin

dust

ries

in20

06.

All

regr

essi

ons

are

wei

ghte

dby

cou

nty

size

.S

tan

dard

erro

rs(i

npa

ren

thes

es)

are

clu

ster

edat

the

stat

ele

vel.

*,**

,an

d**

*de

not

esi

gnifi

can

ceat

the

10%

,5%

,an

d1%

leve

ls,r

espe

ctiv

ely.


these numbers almost perfectly match those obtained at the es-tablishment level. There, the average elasticity of county-levelemployment with respect to house prices in the full LBD samplewas 0.108 versus 0.068 in the matched Compustat-LBD sample(see Table 9 of the Online Appendix).

The second change is that county-level leverage is measuredwith error. Unlike our previous analysis, county-level employmentand county-level leverage are no longer based on the same sampleof firms. While county-level employment now includes all firmsin the LBD, county-level leverage is (still) based on firms in thematched Compustat-LBD sample, because leverage is computedusing data from Compustat. Hence, we only observe Leverage06= Leverage∗

06 + η, where Leverage∗06 is “true” county-level lever-

age based on all firms in the LBD. As is shown in the On-line Appendix, this implies that the coefficients associated withLeverage06 and �Log(HP)06−09 × Leverage06 will be biased down-ward (attenuation bias). By contrast, the coefficient associatedwith the main effect of house prices, �Log(HP)06−09, is biased up-ward and approaches the coefficient from a univariate regressionof �Log(Emp)07−09 on �Log(HP)06−09 from below as σ 2

η → ∞. Inconjunction with the first change discussed above, this impliesthat the coefficient associated with �Log(HP)06−09 increases fortwo reasons: it approaches the average elasticity of county-levelemployment with respect to house prices (i.e., the univariate coef-ficient) from below, while at the same time the average elasticityitself increases because firms become more sensitive to changesin house prices.25

As Table VIII, column (2) shows, all coefficients changein the way discussed above.26 While the coefficient associatedwith �Log(HP)06−09 increases, the coefficients associated withLeverage06 and �Log(HP)06−09 × Leverage06 both decrease (allrelative to the coefficients in column (2) of Table VII). Importantly,

25. Whether the coefficient associated with �Log(HP)06−09 approaches theunivariate coefficient from below or above depends on the sign of the (unbiased)coefficient associated with the interaction term and the sign of the mean valueof county-level leverage. If both are positive, as is the case here, the coefficientassociated with �Log(HP)06−09 approaches the univariate coefficient from below,meaning it increases. See the proof in the Online Appendix for details. Table 20of the Online Appendix simulates measurement error in county-level leverage,empirically validating the results in the proof.

26. Figure 2 of the Online Appendix provides a visual impression similar toFigure III, except that county-level employment includes all firms in the LBD.









the coefficient associated with �Log(HP)06−09 × Leverage06 re-mains large and significant. Thus, despite the downward bias inthis coefficient, we (still) find that variation in county-level lever-age can successfully explain variation in county-level employmentin response to consumer demand shocks, reaffirming that ourprior results are not undone by general equilibrium adjustments.Columns (3) and (4) examine county-level employment in thenontradable sector. The results are similar to those in columns(1) and (2). Finally, columns (5) and (6) confirm prior results thatthere is no significant association between changes in house pricesand changes in county-level employment in the tradable sector.

VII. POLICY IMPLICATIONS

Our results show that job losses during the Great Reces-sion arise from the interaction between declining consumer de-mand and weak firm balance sheets. Mayer, Morrison, and Pisko-rski (2009), Agarwal et al. (2013), and Mian and Sufi (2014b,Chapter 10) discuss policy measures targeted at indebted house-holds. In what follows, we discuss policy measures targeted at(nonfinancial) firms.27

As we discussed in Section IV, financial constraints impairfirms’ ability to engage in labor hoarding. That is, firms with weakbalance sheets cut more jobs in response to a decline in consumerdemand than they would have in the absence of financial con-straints. To explore the role of policy in this context, we wouldlike to draw attention to the case of Germany:

Consider, for a moment, a tale of two countries. Both have suffered asevere recession and lost jobs as a result—but not on the same scale. InCountry A, employment has fallen more than 5 percent, and the unem-ployment rate has more than doubled. In Country B, employment hasfallen only half a percent, and unemployment is only slightly higherthan it was before the crisis. [...] This story isn’t hypothetical. Coun-try A is the United States, where stocks are up, G.D.P. is rising, butthe terrible employment situation just keeps getting worse. CountryB is Germany, which took a hit to its G.D.P. when world trade col-lapsed, but has been remarkably successful at avoiding mass job losses.

27. Our research implies that policy measures could target households, firms,or both. However, if there are diminishing marginal returns to policy intervention,targeting both households and firms is likely to be more effective than targetingeither one alone.


Germany’s jobs miracle hasn’t received much attention in thiscountry—but it’s real, it’s striking, and it raises serious questionsabout whether the U.S. government is doing the right things to fightunemployment (Krugman 2009).

Why did German unemployment barely rise? According tomany commentators, a primary reason is massive labor hoard-ing by German companies (e.g., Dietz, Stops, and Walwei 2010;Balleer et al. 2013; Rinne and Zimmermann 2013).28 Impor-tantly, labor hoarding is heavily subsidized in Germany. A cen-tral pillar of German labor hoarding is the system of short-timework (“Kurzarbeit”) programs encouraging firms to adjust laborthrough hours reductions rather than through layoffs. Under thissystem, firms pay workers for the actual hours worked plus anadditional 60–67% of the net income loss due to the hours reduc-tion. Firms are later reimbursed for this additional pay throughthe unemployment insurance fund administered by the FederalEmployment Agency. Arguably, the success of labor hoarding inGermany may depend on specifics of the German labor market.On a broader level, however, this discussion suggests that it mightbe useful to think about employment policies that target firms di-rectly besides conventional stimulus. As Krugman (2009) argues,“here in America, the philosophy behind jobs policy can be sum-marized as “if you grow it, they will come.” That is, we don’t reallyhave a jobs policy: “we have a G.D.P. policy. [...] Alternatively,or in addition, we could have policies that support private-sectoremployment. Such policies could range from labor rules that dis-courage firing to financial incentives for companies that eitheradd workers or reduce hours to avoid layoffs.”

In the aftermath of the Great Recession, various employmentpolicy measures have been signed into law. For example, the Hir-ing Incentives to Restore Employment (HIRE) Act of 2010 pro-vides tax credits to employers hiring workers who were previ-ously unemployed or only working part time. However, this Actdoes not support the retention of existing workers, which is key tothe idea of labor hoarding. The Layoff Prevention Act—part of theMiddle Class Tax Relief and Job Creation Act of 2012—updatesand clarifies work-sharing provisions in federal law and provides

28. For discussions of Germany’s “labor market miracle” in the Great Reces-sion and its aftermath, see Burda and Hunt (2011), Dustmann et al. (2014), andHoffmann and Lemieux (2016). The latter paper also provides a comparison withCanada and the United States.


federal funding for states to adopt or expand work-sharing pro-grams. Such programs are similar to German-style short-timework programs: employees with reduced hours receive pro-ratedunemployment benefits to supplement their paychecks. However,although work-sharing programs had been in place in 17 U.S.states during the Great Recession, take-up rates were extremelylow.29 According to commentators, the primary reasons are finan-cial disincentives for employers and workers, burdensome filingprocesses, rigid work-sharing schedules, and lack of employer out-reach by state agency officials (e.g., Baker 2011; Abraham andHouseman 2014; Wentworth, McKenna, and Minick 2014).

Last, a word of caution. A potential drawback of German-stylesubsidies is that, while effective in a crisis situation, they mayimpede the efficient allocation of workers in the long run. Laborhoarding may be an optimal response to a temporary decline indemand, but it is not a permanent solution. Accordingly, subsidiesmust be limited in scope and used as temporary relief only.

VIII. CONCLUSION

This article argues that firms’ balance sheets played an im-portant role in the transmission of consumer demand shocks dur-ing the Great Recession. Using establishment-level data from theU.S. Census Bureau, we find that establishments of more highlylevered firms exhibit significantly larger declines in employmentin response to drops in local consumer demand. We find similarresults at the extensive margin: firms with higher leverage atthe onset of the Great Recession are significantly more likely toclose down establishments in response to local consumer demandshocks. These results are not driven by firms being less produc-tive, having expanded too much prior to the Great Recession, orbeing generally more sensitive to fluctuations in either aggregateemployment or house prices. Likewise, at the county level, wefind that counties with more highly levered firms experience sig-nificantly larger job losses in response to county-wide consumerdemand shocks. Thus, firms’ balance sheets also matter for aggre-gate employment. Our research suggests a possible role for em-ployment policies that target firms directly besides conventionalstimulus.

29. Participation in work-sharing programs peaked in 2009 at about 153,000workers, which is just over 0.1% of U.S. payroll employment (Baker 2011).


Our results have implications for macroeconomic modeling.In particular, they suggest that a model in which households’,firms’, and financial intermediaries’ balance sheets interact mightbe a useful way to think about the Great Recession. Accordingly,falling house prices may erode the balance sheets of households,leading to a decline in consumer demand. The latter dispropor-tionately affects firms with weak balance sheets, forcing them todownsize and reduce employment, as is shown in this article. Atthe same time, falling house prices may erode the balance sheetsof financial intermediaries, impairing their capital and access tofunding and therefore their ability to lend. This tightening of lend-ing standards, in turn, disproportionately affects firms with weakbalance sheets (“flight to quality”), reinforcing the adverse effectsof consumer demand shocks.

MIT SLOAN SCHOOL OF MANAGEMENT, NBER, AND CEPRNYU STERN SCHOOL OF BUSINESS, NBER, CEPR, AND ECGI

SUPPLEMENTARY MATERIAL

An Online Appendix for this article can be found at The Quar-terly Journal of Economics online.

REFERENCES

Abraham, Katharine, and Susan Houseman, “Proposal 12: Encouraging WorkSharing to Reduce Unemployment,” in Policies to Address Poverty in Amer-ica, Kearney Melissa and Harris Benjamin, eds. (Washington, DC: BrookingsInstitution, 2014).

Adelino, Manuel, Antoinette Schoar, and Felipe Severino, “House Prices, Col-lateral and Self-Employment,” Journal of Financial Economics, 117 (2015),288–306.

Agarwal, Sumit, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet,Tomasz Piskorski, and Amit Seru, “Policy Intervention in Debt Renegotiation:Evidence from the Home Affordable Modification Program,” Mimeo, Univer-sity of Chicago, 2015.

Aghion, Philippe, Nicholas Bloom, Brian Lucking, Rafaella Sadun, and John vanReenen, “Growth and Decentralization in Bad Times,” Mimeo, Stanford Uni-versity, 2015.

Baker, Dean, “Work Sharing: The Quick Route Back to Full Employment,” Mimeo,Center for Economic and Policy Research, 2011.

Baker, Scott, “Debt and the Consumption Response to Household Income Shocks,”Mimeo, Northwesterm University, 2015.

Balleer, Almut, Britta Gehrke, Wolfgang Lechthaler, and Christian Merkl, “DoesShort-Time Work Save Jobs? A Business Cycle Analysis,” IZA DiscussionPaper 7475, 2013.

Berger, David, and Joseph Vavra, “Consumption Dynamics during Recessions,”Econometrica, 83 (2015), 101–154.




Bernanke, Ben, and Mark Gertler, “Agency Costs, Net Worth, and Business Fluc-tuations,” American Economic Review, 79 (1989), 14–31.

Bernanke, Ben, Mark Gertler, and Simon Gilchrist, “The Financial Accelerator ina Quantitative Business Cycle Framework,” in Handbook of Macroeconomics,Vol. 1, Taylor John and Woodford Michael eds. (Amsterdam: North-Holland,1999).

Biddle, Jeff, “The Cyclical Behavior of Labor Productivity and the Emergence ofthe Labor Hoarding Concept,” Journal of Economic Perspectives, 28 (2014),197–211.

Brav, Alon, Wei Jiang, Frank Partnoy, and Randall Thomas, “Hedge Fund Ac-tivism, Corporate Governance, and Firm Performance,” Journal of Finance,63 (2008), 1729–1775.

Brav, Omer, “Access to Capital, Capital Structure, and the Funding of the Firm,”Journal of Finance, 64 (2009), 263–308.

Brunnermeier, Markus, and Yuliy Sannikov, “A Macroeconomic Model with aFinancial Sector,” American Economic Review, 104 (2014), 379–421.

Burda, Michael, and Jennifer Hunt, “What Explains the German Labor Mar-ket Miracle in the Great Recession?,” Brookings Papers on Economic Activity(2011), 273–319.

Burnside, Craig, Martin Eichenbaum, and Sergio Rebelo, “Labor Hoard-ing and the Business Cycle,” Journal of Political Economy, 101 (1993),245–273.

Caggese, Andrea, and Ander Perez, “The Interaction between Household and FirmDynamics and the Amplification of Financial Shocks,” Mimeo, UniversitatPompeu Fabra, 2016.

Campello, Murillo, John Graham, and Harvey Campbell, “The Real Effects ofFinancial Constraints: Evidence from a Financial Crisis,” Journal of FinancialEconomics, 97 (2010), 470–487.

Charles, Kerwin, Erik Hurst, and Matthew Notowidigdo, “Housing Booms andBusts, Labor Market Opportunities, and College Attendance,” Mimeo, Uni-versity of Chicago, 2015.

Chodorow-Reich, Gabriel, “The Employment Effects of Credit Market Disruptions:Firm-Level Evidence from the 2008–09 Financial Crisis,” Quarterly Journalof Economics, 129 (2014), 1–59.

Clark, Scott, “Labor Hoarding in Durable Goods Industries,” American EconomicReview, 63 (1973), 811–824.

Davis, Steven, “Comment on ‘Job Search, Emotional Well-Being and Job Findingin a Period of Mass Unemployment’ by Alan Krueger and Andreas Mueller,”Brookings Papers on Economic Activity (2011), 58–70.

Davis, Steven, Jason Faberman, and John Haltiwanger, “The Establishment-Level Behavior of Vacancies and Hiring,” Quarterly Journal of Economics,128 (2013), 581–622.

Dietz, Martin, Michael Stops, and Ulrich Walwei, “Safeguarding Jobs throughLabor Hoarding in Germany,” Applied Economics Quarterly, 56 (2010), 125–166.

Dustmann, Christian, Bernd Fitzenberger, Uta Schonberg, and AlexandraSpitz-Oener, “From Sick Man of Europe to Economic Superstar: Ger-many’s Resurgent Economy,” Journal of Economic Perspectives, 28 (2014),167–188.

Eggertsson, Gauti, and Paul Krugman, “Debt, Deleveraging, and the LiquidityTrap: A Fisher-Minsky-Koo Approach,” Quarterly Journal of Economics, 127(2012), 1469–1513.

Fay, Jon, and James Medoff, “Labor and Output over the Business Cycle: SomeDirect Evidence,” American Economic Review, 75 (1985), 638–655.

Foster, Lucia, Cheryl Grim, and John Haltiwanger, “Reallocation in the GreatRecession: Cleansing or Not?,” Journal of Labor Economics, 34 (2016), S293–S331.


Gertler, Mark, and Nobuhiro Kiyotaki, “Financial Intermediation and CreditPolicy in Business Cycle Analysis,” in Handbook of Monetary Economics,Vol. 3, Friedman Benjamin and Woodford Michael, eds. (Amsterdam: North-Holland, 2011).

Gilchrist, Simon, Raphael Schoenle, Jae Sim, and Egon Zakrajsek, “Inflation Dy-namics during the Financial Crisis,” Mimeo, Boston University, 2016.

Gilje, Erik, and Jerome Taillard, “Do Private Firms Invest Differently than PublicFirms? Taking Cues from the Natural Gas Industry,” Mimeo, University ofPennsylvania, 2015.

Guerrieri, Veronica, and Guido Lorenzoni, “Credit Crises, Precautionary Savings,and the Liquidity Trap,” Mimeo, University of Chicago, 2015.

Hall, Robert, “The Long Slump,” American Economic Review, 101 (2011), 431–469.He, Zhiguo, and Arvind Krishnamurthy, “Intermediary Asset Pricing,” American

Economic Review, 103 (2013), 732–770.Hoffmann, Florian, and Thomas Lemieux, “Unemployment in the Great Recession:

A Comparison of Germany, Canada, and the United States,” Journal of LaborEconomics, 34 (2016), S95–S139.

Jarmin, Ron, and Javier Miranda, “The Longitudinal Business Database,” CESWorking Paper 02-17, 2002.

Kaplan, Greg, Kurt Mitman, and Giovanni Violante, “Non-durable Consumptionand Housing Net Worth in the Great Recession: Evidence from Easily Acces-sible Data,” Mimeo, Princeton University, 2016.

Kaplan, Steven, and Luigi Zingales, “Do Financing Constraints Explain Why In-vestment Is Correlated with Cash Flow?,” Quarterly Journal of Economics,112 (1997), 169–215.

Kiyotaki, Nobuhiro, and John Moore, “Credit Cycles,” Journal of Political Economy,105 (1997), 211–248.

Krugman, Paul, “Free to Lose,” New York Times, November 12, 2009.Mayer, Christopher, Edward Morrison, and Tomasz Piskorski, “A New Proposal

for Loan Modifications,” Yale Journal on Regulation, 26 (2009), 417–429.McCue, Kristin, “Matching Compustat Data to the SSEL,” Mimeo, Center for

Economic Studies, 2003.Mian, Atif, Kamalesh Rao, and Amir Sufi, “Household Balance Sheets, Consump-

tion, and the Economic Slump,” Quarterly Journal of Economics, 128 (2013),1–40.

Mian, Atif, and Amir Sufi, “House Prices, Home Equity-Based Borrowing, andthe US Household Leverage Crisis,” American Economic Review, 101 (2011),2132–2156.

——, “What Explains the 2007-2009 Drop in Employment?,” Econometrica, 82(2014a), 2197–2223.

——, House of Debt (Chicago: University of Chicago Press, 2014b).Midrigan, Virgiliu, and Thomas Philippon, “Household Leverage and the Reces-

sion,” Mimeo, New York University, 2016.Moreira, Alan, and Alexi Savov, “The Macroeconomics of Shadow Banking,”

Mimeo, New York University, 2016.Oi, Walter, “Labor as a Quasi-Fixed Factor,” Journal of Political Economy, 70

(1962), 45–66.Rinne, Ulf, and Klaus Zimmermann, “Is Germany the North Star of Labor Market

Policy?,” IMF Economic Review, 61 (2013), 702–739.Rotemberg, Julio, and Lawrence Summers, “Inflexible Prices and Procyclical Pro-

ductivity,” Quarterly Journal of Economics, 105 (1990), 851–874.Sahin, Aysegul, Joseph Song, Giorgio Topa, and Giovanni Violante, “Mismatch

Unemployment,” American Economic Review, 104 (2014), 3529–3564.Saiz, Albert, “The Geographic Determinants of Housing Supply,” Quarterly Jour-

nal of Economics, 125 (2010), 1253–1296.Sharpe, Steven, “Financial Market Imperfections, Firm Leverage, and the Cycli-

cality of Employment,” American Economic Review, 84 (1994), 1060–1074.


Stroebel, Johannes, and Joseph Vavra, “House Prices, Local Demand, and RetailPrices,” Mimeo, New York University, 2015.

Wentworth, George, Claire McKenna, and Lynn Minick, “Lessons Learned: Maxi-mizing the Potential of Work-Sharing in the United States,” Mimeo, NationalEmployment Law Project, 2014.

Whited, Toni, and Guojun Wu, “Financial Constraints Risk,” Review of FinancialStudies, 19 (2006), 531–559.

FIRM LEVERAGE, CONSUMER DEMAND, AND EMPLOYMENT …hmueller/papers/GR.pdf · FIRM LEVERAGE, CONSUMER...

Documents

Transcript of FIRM LEVERAGE, CONSUMER DEMAND, AND EMPLOYMENT …hmueller/papers/GR.pdf · FIRM LEVERAGE, CONSUMER...