answering macro questions using micro data: advances and

Answering Macro Questions Using Micro Data:Advances and Challenges

Maximilian Eber and Xavier Jaravel

March 26, 2014

Maximilian Eber and Xavier Jaravel Answering Macro Questions Using Micro Data: Advances and ChallengesMarch 26, 2014 1 / 92

1 Introduction

2 Methodology

3 Data Sources

4 Examples from Different Areas

5 Jimenez et al. (2014)

6 Mian and Sufi (2011)

7 Gourinchas and Parker (2002)

8 Future Research


Section 1

Introduction


Introduction

From Chetty (2012)


Introduction

From Klenow (2008):

I found Banerjee’s paper a pleasure to read, both provocative and compelling. As

Banerjee concedes, the most important questions in development are often “macro” ones

such as how to boost growth in the fashion of China and India in recent decades. But,

Banerjee emphasizes, we don’t know very much about how to do this despite lots of

research effort. So he suggests we largely abandon macro development research and

concentrate on micro development questions such as the effect of randomized

interventions in health and education. The following Table summarizes Banerjee’s

position.

Welfare Welfare Knowledge = Research Knowledge Research

Micro 0 Low High

Macro 0? High Low

w w wx

w w w

!

|

I wholeheartedly agree with the “micro” row in this Table, so will say no more

about it. Instead I will defend the value of doing macro development research. As I

consider myself at the micro data end of the macro spectrum, I do this without feeling

particularly defensive.

I begin by drawing an analogy to potential micro vs. macro divides in medicine

and biology. Consider “micro” researchers in medicine who do carefully controlled

clinical trials and lab work. I imagine they might disdain “macro” medicine, which looks

at trends in life expectancy, heart disease, cancer, autism, etc. Their skepticism may


Macro Questions Meet Micro Data

Which papers qualify?I Must contribute to our understanding of key economy-wide variablesI Papers vary along various dimensions

F General vs. Partial EquilibriumF Quasi-experimental variation (e.g. IV) vs. Model-based identification

(e.g. structural work)F High-quality micro data (e.g. universe of firms) vs. Disaggregated data

(e.g. large sample of firms)


Plan for today

1 Introduction

2 Methodology

3 Data Sources





8 Future Research


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 2

Methodology


Quasi-Experimental Research Designs

Key: finding exogenous variationStrengths

I Transparent and credible identification

WeaknessesI Does not identify policy-invariant parameters and has limited relevance

for policy and welfare analysis

Mian and Sufi (2011) & Jimenez et al. (2014)


Structural Approach

Key: having a good modelStrengths

I Ability to make predictions about counterfactual outcomes and welfareI Addresses GE effects

WeaknessesI Difficult to credibly identify all primitive parameters (because of

selection effects, simultaneity bias and omitted variables)I Model uncertainty (robustness to small changes in assumptions)

Parker and Gourinchas (2002)


Sufficient Statistic Approach

StrengthsI Formulas for the welfare consequences of various policies are derived as

functions of reduced-form elasticities rather than structural primitivesI Less data and variation are needed than to fully identify a structural

model

WeaknessesI A new sufficient statistic formula must be derived for each questionI More easily misapplied than structural method

Chetty (2009), Shimer and Werning (2008).


Example: the Deadweight Loss of TaxationWorkers solve

maxl u(c, l) = c−ψ(l) s.t. c = y + (1− t)wl

Welfare

W (t) = {y + (1− t)wl−ψ(l)}+ twl

Therefore,

dW (t)

dt =dudt +

d [twl ]dt

=∂ u∂ t + wl + t dwl

dt

= 0 + t dwldt

= t dTIdt


Challenges compared to “Micro”

General EquilibriumI Prices change, technology responds, etc...I What is the control group?

AggregationI It is sometimes hard to map macro model (lacking heterogeneity) to

micro data (e.g. CCAPM)


Example: Dealing with GE

Adjustment Costs, Firm Responses, and Micro vs. Macro LaborSupply Elasticities: Evidence from Danish Tax Records (Raj Chetty,John Friedman, Tore Olsen, and Luigi Pistaferri), Quarterly Journal ofEconomics 126(2): 749-804, 2011


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 3

Data Sources


Compustat

Very popularFazzari, S., Hubbard, R., & Petersen, B. (1988). FinancingConstraints and Corporate Investment. Brookings Papers onEconomic Activity.Graham, J. R. (2000). How Big Are the Tax Benefits of Debt?Journal of Finance, 55(5), 1901–1941.


Surveys

Murillo Campello & Erasmo Giambona & John R. Graham &Campbell R. Harvey, 2011. ”Liquidity Management and CorporateInvestment During a Financial Crisis,” Review of Financial Studies,Society for Financial Studies, vol. 24(6), pages 1944-1979.


Census Data

Parker, J. A., Souleles, N. S., Johnson, D. S., & McClelland, R.(2013). Consumer Spending and the Economic Stimulus Payments of2008. American Economic Review, 103(6), 2530–2553.


Tax Records: Scandinavia

Laurent E. Calvet & John Y. Campbell & Paolo Sodini, 2009.”Measuring the Financial Sophistication of Households,” AmericanEconomic Review, American Economic Association, vol. 99(2), pages393-98, May.Laurent E. Calvet & John Y. Campbell & Paolo Sodini, 2007. ”Downor Out: Assessing the Welfare Costs of Household InvestmentMistakes,” Journal of Political Economy, University of Chicago Press,vol. 115(5), pages 707-747, OctoberLaurent E. Calvet & John Y. Campbell & Paolo Sodini, 2009. ”FightOr Flight? Portfolio Rebalancing by Individual Investors,” TheQuarterly Journal of Economics, MIT Press, vol. 124(1), pages301-348, February.


Financial Regulation 1: Credit Registers

Gabriel Jimenez & Steven Ongena & Jose-Luis Peydro & JesusSaurina (2012). ”Credit Supply and Monetary Policy: Identifying theBank Balance-Sheet Channel with Loan Applications,” AmericanEconomic Review, pages 2301-26


Financial Regulation 2: Syndicated Loan Market

Chodorow-Reich, Gabriel (2014). The Employment Effects of CreditMarket Disruptions: Firm-level Evidence from the 2008-09 FinancialCrisis. Quarterly Journal of Economics129, no. 1: 1-59.


Internet (”Scraping”)Cavallo A, B Neiman and R Rigobon (2013), “Currency Unions,Product Introductions, and the Real Exchange Rate”, QuarterlyJournal of Economics, forthcoming.


Patent Applications from USPTO

Identifying Technology Spillovers and Product Market Rivalry,Nicholas Bloom, Mark Schankerman and John Van Reenen,Econometrica (2013)Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith andPeter Howitt (2005). ”Competition And Innovation: An Inverted-URelationship,” Quarterly Journal of Economics, 701-728.More on this with John Van Reenen on April 9th


More Data

Consumer credit bureau dataHouse price indicesMortgage application HMDA dataMortgage dataBank and Stocks dataEmployment and Payroll dataAmerican Community SurveyIncome DataConsumptionPricesSee handout for details


Better Data

Still a lot to wish for...Universe of transactionsFirm Side:

I High-frequency data on firm balance sheets, prices andcustomer-supplier networks

I Trace business-cycle shocks through the economy

Consumers:I High-frequency consumption, income and asset data


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 4

Examples from Different Areas


Growth and Productivity

Chang-Tai Hsieh & Peter J. Klenow, 2009. ”Misallocation andManufacturing TFP in China and India,” The Quarterly Journal ofEconomics, MIT Press, vol. 124(4), pages 1403-1448.Nicholas Bloom & Benn Eifert & Aprajit Mahajan & David McKenzie& John Roberts, 2013. ”Does Management Matter? Evidence fromIndia,” The Quarterly Journal of Economics, Oxford University Press,vol. 128(1), pages 1-51.


Local Multipliers

State levelI Clemens and Miran (2010), Cohen et al. (2010), Serrato and

Wingender (2010), Shoag (2010), Acconcia et al. (2011),Chodorow-Reich et al. (2011), Fishback and Kachanovskaya (2010),Nakamura and Steinsson (2013).

Firm levelI Mendel (2013)


Banking

Rajkamal Iyer, Samuel Lopes, Jose-Luis Peydro and AntoinetteSchoar (2014). ”The Interbank Liquidity Crunch and the Firm CreditCrunch: Evidence from the 2007-09 Crisis”, forthcoming in Review ofFinancial Studies


Risk-Channel of Monetary Policy

Gabriel Jimenez & Steven Ongena & Jose Luis Peydro & JesusSaurina (2014). ”Hazardous Times for Monetary Policy: What do 23Million Loans Say About the Impact of Monetary Policy on CreditRisk-Taking?”, forthcoming in Econometrica.


Real Effects of Financial Frictions

Chodorow-Reich, Gabriel. 2014. The Employment Effects of CreditMarket Disruptions: Firm-level Evidence from the 2008-09 FinancialCrisis. Quarterly Journal of Economics129, no. 1: 1-59.Rauh, Joshua D, ”Investment and Financing Constraints: Evidencefrom the Funding of Corporate Pension Plans”, Journal of Finance 61,1 (2006), pp. 33–71.


Inflation

Peter J. Klenow & Oleksiy Kryvtsov, 2008. ”State-Dependent orTime-Dependent Pricing: Does It Matter for Recent U.S. Inflation?,”The Quarterly Journal of Economics, MIT Press, vol. 123(3), pages863-904, August.Klenow, Peter J. & Malin, Benjamin A., 2010. ”MicroeconomicEvidence on Price-Setting,” Handbook of Monetary Economics, in:Benjamin M. Friedman & Michael Woodford (ed.), Handbook ofMonetary Economics, edition 1, volume 3, chapter 6, pages 231-284Elsevier.Mark Bils & Peter J. Klenow & Benjamin A. Malin, 2012. ”ResetPrice Inflation and the Impact of Monetary Policy Shocks,” AmericanEconomic Review, American Economic Association, vol. 102(6),pages 2798-2825, October.


International Macro

Gopinath, Gita, Oleg Itskhoki, and Roberto Rigobon. “CurrencyChoice and Exchange Rate Pass-through.” American EconomicReview 100(1) (2010): 304–36.


Housing Market

Mian and Sufi’s research agenda:I The Consequences of Mortgage Credit Expansion: Evidence from the

U.S. Mortgage Default Crisis (QJE 2009)I House Prices, Home Equity-Based Borrowing, and the U.S. Household

Leverage Crisis (AER 2011)I Household Balance Sheets, Consumption, and the Economic Slump

(QJE 2013)

Nathanson, Charles G, and Eric Zwick. Working Paper. ArrestedDevelopment: Theory and Evidence of Supply-Side Speculation.


Expectations and Learning

Ulrike Malmendier & Stefan Nagel (2011). ”Depression Babies: DoMacroeconomic Experiences Affect Risk Taking?,” The QuarterlyJournal of Economics, Oxford University Press, vol. 126(1), pages373-416.Ulrike Malmendier, S. Nagel (2014), “Learning from InflationExperiences”, Quarterly Journal of Economics, Forthcoming


Business Cycles

Xavier Gabaix: ”The Granular Origins of Aggregate Fluctuations,”Econometrica, 2011 vol. 79, p.733-772


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 5

Jimenez et al. (2014)


Risk-Channel of Monetary Policy

Gabriel Jimenez, Steven Ongena, Jose Luis Peydro, and Jesus Saurina(2009): ”Hazardous times for monetary policy: What do twenty-threemillion bank loans say about the effects of monetary policy on creditrisk-taking?”, forthcoming in Econometrica.


TheoryAgency Conflicts and Risk-Taking

Story they want to tell:I Central bank lowers rates...I ... and banks with agency conflicts...I ... take on excessive risk.

Story they actually tell:I Central bank lowers rates...I ... and banks with low capital ratios...I ... lend more to risky borrowers.


Data

Spanish credit registerI all loan applications and all loans granted in Spain (above 6,000 Euros)I 1984 to 2009, includes pre-Euro and post-Euro phase and run-up and

impact of Great Recession

“Fairly exogenous” interest rateI 97% variation of Spanish rate explained by German rateI GDP correlation only 36%.I Spanish GDP only 15% of Eurozone.


Main SpecificationDo banks originate more loans to risky borrowers when rates decrease?

Specification:

∆ln (LOANbit) =α + β ∆IRt−1 + γ ∆IRt−1× ln (CAPITALb,t−1)

+ δ ∆IRt−1× ln (CAPITALb,t−1)× I(DOUBTFULi ,t)

+ CONTROLSb,i ,t + εb,i ,t

δ > 0: More lending to risky borrowers by weakly capitalized banks

∆IR ↓,CAPITAL ↓,δ > 0⇒∆(LOANb,i ,t) ↑


Economic Significance

Weakly capitalized banks (Q25, 4.26%) grant 25% more credit torisky borrowers, relative to strongly capitalized banks (Q75, 7.2%),when the interest rate declines by a standard deviation (1.6pp).


Discussion and Extensions

AdvantagesI Nice datasetI Robust empirics

Disadvantages:I Channels?I Welfare?I Policy implications?


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 6

Mian and Sufi (2011)


House Prices, Home Equity–Based Borrowing, and the USHousehold Leverage Crisis

What is the causal effect of house price growth on homeownerborrowing?

I US household leverage sharply increased between 2002 and 2007I Wider availability of mortgage credit for riskier home buyers, driving up

house prices (Mian and Sufi (2009))I Feedback: how did homeowner borrowing respond to the increase in

house prices?


The Home-Equity Based Borrowing Channel

Theory:I unconstrained long-lived homeowners perfectly hedged against house

price fluctuations: propensity to borrow out of housing is zero.I short-lived homeowners: borrow against unexpected increases in home

equity to finance consumption, with the propensity to borrow beingstronger for older homeowners.

I credit-constrained homeowners.I homeowners with limited self-control.


The Home-Equity Based Borrowing Channel

Changes in house prices and homeowner borrowing are jointlydetermined (e.g. shock to expected income growth)Need exogneous source of variation in house price growth


Preview of IV Strategy and Results

Across-MSA instrument: housing supply elasticity at the MSA levelfrom Saiz (QJE 2010).

I Estimate: borrowing of 25 cents for every dollar gain in home equityfrom 2002 to 2006.

Within-MSA instrument: fraction of subprime borrowers in a zip codeinteracted with MSA elasticityHeterogeneity across households:

I high vs. low credit card utilization rates; high vs. low initial creditscores

I young vs. old


Data

Credit report information from EquifaxI 74,149 homeowners in 2,307 zip codes located in 68 MSA

This “individual” data is augmented with several additional datasetsat the zip-code level: income, house prices, credit scores, ...


Across-MSA Empirical StrategyIntuition: for an equivalent housing demand shock, the slope of thehousing supply curve determines the degree to which housing pricesrise in an area (Glaeser, Gyourko, and Saiz (2008)).


Across-MSA TSLS

LeverageGrowth0206izm = θXizm + β HousePriceGrowth0206zm + vizm

HousePriceGrowth0206zm = δ Xizm + ρElasticitym,1997 + εizm

Individual i lives in zipcode z within MSA mSource of variation is ultimately at the MSA level.Second-stage estimates suggest that 1997 homeowners borrow 25cents on every dollar of additional home equity.


Across-MSA TSLS


Cross-Sectional Heterogeneity

Also serves as a Placebo test for home-equity based borrowingchannel.


Exclusion Restriction


Further Robustness Checks

Within-MSA estimation technique exploiting variation at the zip codelevel.Builds on Mian and Sufi (2009)

I from 2002 to 2005 house price appreciation in subprime zip codes wassignificantly stronger than in prime zip codes in the same MSA.

I this holds only in MSAs with inelastic supply of housingI driven by an expansion in the supply of credit to subprime zip codes

everywhere.


Macroeconomic Impact

What do people do with the borrowed money?I Not used for real estate or financial assetsI Not use to pay down credit card debt

Home Equity-Based Borrowing and DefaultsI From 2006 to 2008, homeowners in inelastic MSAs experience a sharp

increase in default ratesI Effect comes from low-credit-quality borrowers


Macroeconomic Impact

Claim 1: from 2002 to 2006, house price appreciation causedaggregate borrowing to increase by $1.25 trillion dollars, which is 53%of the overall increase in debt of homeowners during this period.Claim 2: defaults due to the home equity-based borrowing channelrepresent 39% of total new defaults in the US economy after 2006.LATE vs. GE effects?


Discussion and Extensions

Nice datasetSimple instruments, exposition and Placebo testsAggregation procedure and macroeconomic implications unclear

I Bernanke-Gertler (1989), Kiyotaki-Moore (1997)I Eggertsson-Krugman (2012)


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 7

Gourinchas and Parker (2002)


Empirical Profiles (by Education)

68 p.-o. gourinchas and j. a. parker

Figures 3 and 4 give some evidence that consumption and income track eachother across subgroups of the population defined by education and occupationgroups. These graphs are somewhat noisy. However, despite the noise, one cansee that the occupation and education groups with the most pronounced humpsin income present the most pronounced humps in consumption. Further, we canformally reject the null hypothesis that the consumption profiles are flat. This isessentially a now standard test of the linearized consumption Euler equation, asstudied by Attanasio and Weber (1995), Lusardi (1996).Our profiles differ slightly from the results of Attanasio and Browning (1995)

and Attanasio and Weber (1995). These papers employ a larger set of preferenceshifters: once controlling for these, consumption is smoother and the CEQ-LCHis not rejected. In Attanasio and Weber (1995) and in the linear Euler-equationapproach generally used in micro data, precautionary effects are omitted so thatpreference shifters absorb, correctly or incorrectly, variation in consumption thatwe attribute to uncertainty. Clearly, allowing for enough preference variation can

Figure 3.—Household consumption and income over the life cycle, by education group.


Empirical Profiles (by Profession)consumption over the life cycle 69

Figure 4.—Household consumption and income over the life cycle, by occupation group.

explain the life-cycle correlation without reference to substantial household-leveluncertainty. The approach of this paper is to adjust only for the size of the house-hold and examine how precautionary saving explains the observed consumptionbehavior. Our methodology has the potential to nest and test both explanations.It is also worth noting that our profiles represent total consumption expendituresrather than nondurable consumption.

6! results

We first estimate our model for the average household and discuss the impli-cations of the fitted structural model for household behavior. We then turn todisaggregated results, by education and occupation groups, and to evaluating therobustness of our findings.

6!1! Findings from the Entire Sample

As a useful benchmark, we begin by asking what the standard Life-Cycle theorywould predict, assuming away all uncertainty. To give the best chance to theCEQ-LCH, we perform first difference estimation, not asking it to fit the meanof the consumption profile. Under certainty, equation (2) holds, implying, after


Data

Consumer Expenditure Survey (CEX)I 40,000 households, 1980-1993I construct consumption and income patterns over the life-cycleI average income and consumption by age

Panel Study of Income Dynamics (PSID)I income by household and ageI used to estimate income process


Structural Approach

1 Write down a model of consumption over the life cycle2 Use the data to generate a path of state variables.3 Simulate the model over this path4 Compare the resulting “artificial” consumption profile to the data.5 Choose model parameters to maximize the fit


Model Setup (Simplified)

Objective function: consumption and bequests.Maximize

E0

[N

∑t=1

βtu (Ct) + β

N+1VN+1 (WN+1)

]subject to the law of motion for wealth and (self-imposed) borrowingconstraint.Parameterize u(.) and V (.) as CES and estimate those parameters.


Consumption Rule56 p.-o. gourinchas and j. a. parker

Figure 1.—A tale of two households.

age 66 characterized by !0 = 0"001#!1 = 0"071. These parameter values are thoseof our baseline estimates.15 If G and Z are constant, the finite horizon problemwould converge to the infinite horizon one, as we move further away from retire-ment. Consumption is always positive, increasing, and concave in cash on hand.Early in life, households exhibit ‘buffer stock’ behavior: for low levels of cashon hand, typically less than the permanent component of their income $x ≤ 1%,households consume most, but never all, of their financial wealth, and move tothe next period with a very low level of liquid assets. At high levels of cash onhand, the precautionary motive is small and households consume more than theincome they expect to receive (which equals 1) and so run down their assets. Aswe discuss later, our baseline retirement consumption rule implies either littleilliquid wealth or a low propensity to consume from it. Thus as households age,they must save for retirement: consumption rules decline and households accu-mulate significant amounts of liquid wealth.

15 Other relevant parameters are &= 0"960#'= 0"514#R= 1"0344, typical income uncertainty, fam-ily size, and expected income profile, as discussed in Sections 4 and 5.


Estimating the Model

Take a set of parameters ψ, which generates the consumption policyfunction for individual i is

Ci ,t = Ct(si ,t ;ψ) = ct(xi ,t ;ψ)Pi ,t

which maps from (some observed) state variables si ,t to predictedconsumption.Postulate

ln (Ci ,t) = ln (ct(xi ,t ;ψ)Pi ,t) + εi ,t

and use a minimum distance estimator (GMM).


Method of Simulated Moments

Previous approach requires high quality data on individuals’ statevariables and consumption choices.

I We do not have those.

Method of Simulated Moments allows to use more coarse dataI Bring the model close to any moment of the data

See Pakes and Pollard (1989) for methodology


Model Fit (Full Sample)72 p.-o. gourinchas and j. a. parker

Figure 5.—The fitted consumption profile.

alternative interpretation for a low !0 is that households have a low propensityto consume out of illiquid wealth at retirement.34It should be noted that both estimation methods reject the overidentifying

restrictions at the 5% level. The 95% critical value for a "2#36$ is 50%71 and thechi-square always exceeds 150. This is not entirely surprising, given the numberof moments we use (40) and the few parameters of the model. The estimatedmodel should still be taken seriously however. As we now discuss, the modeldoes much better in an economic sense than the CEQ-LCH model with whichthis section begins.With our estimates in hand, we can address how well the stochastic model fits

the life-cycle consumption profile. The first panel of Figure 5 plots the simulatedand actual consumption data along with the income profile. The stochastic life-cycle model does a much better job at fitting the consumption profile than theconsumption profile with constant growth rate of #1/&$ ln#'R$ that would obtainunder the certainty-equivalent. The consumption profile from the fitted model

34 For instance, if the ratio of illiquid wealth to permanent income, h, were equal to 6, the marginalpropensity to consume out of illiquid wealth would be a mere 0%35%.


Counterfactuals: Savings Distribution

consumption over the life cycle 75

life-cycle consumption:

SLCt = !WLC

t+1 −WLCt "/R= !R−1"/RWLC

t +YLCt −CLC

t #

Precautionary saving is the complement of life-cycle saving.The first panel of Figure 7 plots the precautionary and life-cycle liquid saving

of the average household. Given the estimated discount rate and the profile ofexpected income, young consumers facing no income risk would like to borrowlarge amounts, so life-cycle saving is negative early in life. Young households infact hold a positive buffer stock of wealth in response to income risk, so that pre-cautionary saving is positive early in life. In the early to mid forties, in accordancewith our previous discussion, life-cycle saving becomes larger than precautionarysaving. Households begin to build their liquid wealth for retirement purposes. Asasset levels increase, the expected variance of consumption declines, decreasingthe precautionary saving motive. Since households that face income uncertaintysave more early in life due to risk, they are able to consume more and save lesswhen older, leading to negative precautionary saving late in life. This discussion

Figure 7.—The role of risk in saving and wealth accumulation.


Summary: Structural Approach

ProsI Counterfactuals and Policy analysisI Flexible data usage

ConsI Model uncertaintyI Computational complexity


1 Introduction

2 Methodology

3 Data Sources





8 Future Research


Section 8

Future Research


Empirical Work to Inform Macroprudential Regulation

Aggregate Demand ExternalitiesPecuniary externalities (Fire sales etc.)Financial regulation

I Structural model / sufficient statistic for optimal interventions (e.g.liquidity requirements)


Aggregate Demand Externalities

Renewed interest in the theory of aggregate demand externalitiesI Farhi and Werning (2012)I Farhi and Werning (2013)

Can we assess empirically whether aggregate demand externalitiesexist and their quantitative importance?


Empirical Strategy

Consider exogenous shocks hitting US countiesWhat are the implications of a model with aggregate demandexternalities for the cross-sectional distribution of job losses?

I there will be more job losses in counties that are more “closed”I these job losses will be concentrated in the non-tradeable sectorI similar in spirit to Mian and Sufi (2012)I openness of the local economy could be measured by the steady-state

employment share of non-tradeables

Exogenous shocksI deleveraging shocks (as in Mian and Sufi (2012))I fiscal transfers (as in Wilson Chodrow-Reich et al. (2012) and Wilson

(2012))I military expenditures (as in Nakumura and Steinsson (2013))


Empirical Specification

Borrow the model form Farhi and Werning (2012) to characterize theresponse of the economy to a “transfer” in a dynamic model

βc,t = evt

ρ1−α

α− (1− evt)ρ

11

1−χ+ φ

I estimate the following model:

Lt+1,i −Lt,iLt,i

= γ1St,i + γ2St,i · It,i + β Xt,i + ut,i

t indexes time and i the local economy. Li is employment, Si anexogenous shock to the local economy (think of predicted value inTSLS framework) and Ii the degree of openness of the local economy.Xt,i is a rich set of covariates.


Nontradeable Employment Shares


Deleveraging Shock


Fiscal Policy Shock


To Be Continued?

Better measure of nontradeablesBetter measure of openness of the local economy (production vs.consumption)Explanation for the cross-sectional variation in non-tradeableemployment shares

I Instrument for non-tradeable employment shares?


Macro and Networks

Daron Acemoglu & Vasco M. Carvalho & Asuman Ozdaglar & AlirezaTahbaz Salehi, 2012. ”The Network Origins of AggregateFluctuations,” Econometrica, Econometric Society, vol. 80(5), pages1977-2016, 09.Engelberg, J., Gao, P., and Parsons, C., 2012, Friends with Money,Journal of Financial Economics, 103 (1), p. 169–188.Possible directions:

I studying contagion in financial networks using Fed data


answering macro questions using micro data: advances and

Documents