《IMF-异构支出异构乘数（英）-2023.3.pdf》由会员分享，可在线阅读，更多相关《IMF-异构支出异构乘数（英）-2023.3.pdf（51页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、Heterogeneous Spending,Heterogeneous Multipliers Umberto Muratori,Pedro Juarros and Daniel Valderrama WP/23/WP/23/52 IMF Working Papers describe research in progress by the author(s)and are published to elicit comments and to encourage debate.The views expressed in IMF Working Papers are those of th
2、e author(s)and do not necessarily represent the views of the IMF,its Executive Board,or IMF management.2023 MAR*Umberto Muratori,European University Institute;Pedro Juarros,International Monetary Fund;Daniel Valderrama,The WorldBank.2023 International Monetary Fund WP/23/52IMF Working Paper Western
3、Hemisphere Department Heterogeneous Spending,Heterogeneous Multipliers Prepared by Umberto Muratori,Pedro Juarros and Daniel Valderrama*Authorized for distribution by Manuela Goretti March 2023IMF Working Papers describe research in progress by the author(s)and are published to elicit comments and t
4、o encourage debate.The views expressed in IMF Working Papers are those of the author(s)and do not necessarily represent the views of the IMF,its Executive Board,or IMF management.ABSTRACT:Do local fiscal multipliers depend on what the government purchases?We find that government purchases of service
5、s have larger effects on employment than spending on goods.Industries producing services are more labor-intensive than industries producing goods.This heterogeneity in labor intensity is an important mechanism behind these results.Spending directed toward labor-intensive industries generates stronge
6、r increases in employment and labor income relative to spending toward non-labor-intensive industries.JEL Classification Numbers:E10,E62,H57,R12 Keywords:Heterogeneous Local Fiscal Multipliers;Services and GoodsSpending;Labor IntensityAuthors E-Mail Address:Pedro Juarros:Pjuarrosimf.org;Umberto Mura
7、tori:umberto.muratorieui.eu;Daniel Valderrama:dvalderramagonzaworldbank.org IMF WORKING PAPERS Heterogeneous Spending,Heterogeneous Multipliers INTERNATIONAL MONETARY FUND 3 WORKING PAPERS Heterogeneous Spending,Heterogeneous Multipliers Prepared by Umberto Muratori,Pedro Juarros and Daniel Valderra
8、ma 1 1 We are grateful to Laurent Bouton,Russell Cooper,Mark Huggett,Toshihiko Mukoyama,Myrto Oikonomou,Yongquan Cao,and Martin Ravallion for their helpful comments.We also thank seminar participants at the University of Florence and the University of Siena and conference participants at the Max-Web
9、er Programme June Conference.We thank Digna Isabel Andrade Portilla for her excellent research assistance.We acknowledge the generous financial support of the Washington Center for Equitable Growth.The views in this paper are solely those of the authors and should not be interpreted as reflecting th
10、e views of the International Monetary Fund or World Bank.All errors are our own.1IntroductionThe empirical literature on the government purchase multiplier commonly treats government spendingas a homogeneous good.Very little is known about the heterogeneity in the fiscal multipliers comingfrom the h
11、eterogeneity in spending(Chodorow-Reich,2019),mainly due to the lack of granular data ongovernment purchases.The research on local fiscal multipliers has used variation in public spending comingfrom either the American Recovery and Reinvestment Act(Chodorow-Reich et al.,2012;Wilson,2012;Conley and D
12、upor,2013;Dupor and Mehkari,2016)or Department of Defense(DOD)purchases(Nakamuraand Steinsson,2014;Dupor and Guerrero,2017;Demyanyk et al.,2019;Auerbach et al.,2020)across regionsin the US.These studies estimate local fiscal multipliers using aggregate spending and do not capture thevariation in the
13、 estimates associated with different types of spending.Our paper fills this gap by answering two questions.Do different types of government purchases generatedifferent multipliers?What are the mechanisms driving this heterogeneity?We assemble a unique datasetcontaining over 40 years of military cont
14、ract-level US procurement spending at an annual frequency.1Ourdata report information about the products that contracts require recipients to produce and the locationwhere a contract is performed.We leverage this new information to classify fiscal spending into two categories:purchases of goods and
15、purchases of services.The geolocation of contracts at the metropolitan statisticalarea(MSAs)in the US allows quantifying the causal effects of types of public purchases on employmentusing the cross-sectional and time variation(Nakamura and Steinsson,2014).The breakdown into goods versus services is
16、a natural distinction to explore.The governments coredecision is the budget allocation between these two types of products,for which it demands input factorsto provide public goods and services to its clients.Our paper shows that changing the budget allocationbetween goods and services can improve t
17、he effectiveness of fiscal packages.Goods and services differ alongseveral dimensions,such as labor intensity,tradability,and productivity.Their comparison sheds light onthe relevance of specific channels in generating positive aggregate responses.Exploring these mechanismsleads to a better understa
18、nding of the effectiveness and design of fiscal stimuli.2The first novel result of our paper consists of the switch of government defense spending from goods toservices starting in the early 1990s.Before that date,more than 65%of procurement defense spending wasused to purchase goods.In recent decad
19、es,the US federal government has redesigned its budget allocation,purchasing approximately equal shares of goods and services.1For the analysis of this project,we rely on defense spending.Defense spending composes the largest fraction of federalprocurement spending;it accounts for about 50%of transa
20、ctions by count and more than 60%by value.Cox et al.(2020)show that the characteristics of defense spending are similar to those of the spending made by other agencies.We are unlikelyto miss important features by restricting our attention to defense spending.Furthermore,due to political accountabili
21、ty,theUS federal government has disclosed defense procurement contracts for longer than contracts signed by other federal agencies,allowing us to leverage a long time series for the analysis.Finally,as argued by Nakamura and Steinsson(2014),variation inspending awarded by the Department of Defense(D
22、oD)is mainly driven by geopolitical events,and it is less sensitive to localeconomic conditions than the spending awarded by other federal agencies,making our identification strategy stronger.2Ramey and Shapiro(1998)show that changes in the composition of government spending across sectors generate
23、negativeresponses in the aggregate variables due to costly reallocation of the production factors.Cox et al.(2020)highlight that thedegree of sectoral price stickiness impacts the fiscal transmission mechanism.2We document significant differences in the employment of local fiscal multipliers between
24、 shocks tospending in goods or services.A one-percentage-point increase in government services spending(normalizedby personal income)generates an on-impact employment response of 0.36 percent.The estimates increase astime passes and become 1.19 two years after the shock.3Goods spending employment mu
25、ltipliers are positivebut non-significant at any horizon after the occurrence of the shock.The estimates for goods spending rangebetween 0.06 percent at impact and 0.19 percent at the two-year horizon,respectively;they are betweenone-fourth and one-tenth smaller than the estimates for services spend
26、ing.Unlike goods spending,servicesspending also leads to a significant increase in the households labor income.To grasp a better sense of thesize of these estimates,we calculate the newly-created jobs for$100,000 spent in procurement contracts.Thenumber of new jobs created substantially differs betw
27、een the two types of spending.$100,000 in procurementcontracts that acquire services create 0.81 new jobs at impact,and 2.71 after two years.By contrast,thepurchase of goods generates 0.13 new jobs at impact and 0.36 after two years.Chodorow-Reich(2019)reports a range from 0.76 to 3.80 of new jobs c
28、reated at the two-year horizon.The number of new jobscreated by services spending is close to the upper bound of that range,while the new jobs generated bygoods spending is slightly below the lower bound.Our estimates are in line with the literature and highlightthe role of heterogeneous spending on
29、 the size of the local fiscal multiplier.We also investigate whether the employment responses occur in the industries that receive the procure-ment spending or the effects spill over to industries that do not receive the spending.We find that bothtypes of spending generate positive and significant d
30、irect effects on the industry that receives most of thatspending.4Still,indirect effects consisting of positive employment responses that spill to other sectors occuronly after service spending shocks.We find that labor income increases only after services spending shocks,generating multiplier effec
31、ts that propagate to all industries.5These results highlight the importance of thetransmission of fiscal spending through changes in labor income.In the second part of the paper,we separately test three mechanisms that may explain the differences inthe estimates between services and goods spending.F
32、irst,the production of services is more labor-intensivethan the production of goods.Government dollars spent on acquiring services may increase the demandfor workers,labor income,and private consumption by more than the dollars spent in producing goods.This pass-through mechanism may generate a virt
33、uous circle leading to larger multiplier effects after servicesspending shocks.Second,goods are more tradeable than services.If the production of goods spills intoneighboring locations,small fiscal multiplier effects after goods spending shocks may be due to geographicspillovers.Third,government spe
34、nding may generate demand-driven growth by facilitating firm turnover3Two-year fiscal multipliers are the benchmark estimates in the literature for three reasons.First,two years is the policy-relevant time frame for counter-cyclical policy packages.Second,it reduces the measurement error that comes
35、from differencesbetween the fiscal year,the unit of measurement of public spending,and the calendar year,the unit of measurement of economicactivity.Third,the bias of not adopting a fully dynamic specification fades away as time elapses.4About 80%of the goods spending is allocated to contractors bel
36、onging to the manufacturing sector,and 55%of the servicesspending is directed to contractors in the services industry.5Alonso(2017)and Bouakez et al.(2020)show that the direct effect of government spending shocks is relatively small andthat most aggregate impact comes from the indirect effect of hou
37、seholds responses to increases in their income.3or improving innovation.If firms producing services differ from those producing goods,a government shockmay affect their productivity gains differently and generate heterogeneous multiplier effects.Labor intensity is an important factor in driving the
38、differences by spending category.Industries thatproduce services are more labor-intensive than industries that produce goods.We divide each spendingcategory into labor-intensive and non-labor-intensive sub-components,and we estimate the employmentmultipliers for the new sub-components.On the one han
39、d,the effects of spending in non-labor-intensiveindustries are negligible or even negative.On the other hand,shocks to either type of spending directed tolabor-intensive industries generate positive and significant increases in employment.By contrast,we do not find strong support for tradeability an
40、d productivity gains as the main driversof the differences in the local multipliers by type of spending.We implement two strategies to test thecontribution of tradeability and geographic spillovers in explaining the differences in the estimates betweenservices and goods spending.The first test consi
41、sts in replicating the analysis at a higher geographicaggregation at state levelin which the geographic misallocation of procurement contracts is less relevant.The second test quantifies the“outflow”effects on neighboring locations.Both tests indicate that geographicspillovers do not explain the dif
42、ferences in the multipliers by category of spending.At the same time,we explore if the two types of spending shocks have different impacts on firms entryrates,exit rates,or innovation outcomes.Our tests suggest that shocks to services spending increase firmsentry rates,decrease exit rates,and hurt i
43、nnovation activities.Shocks to goods spending only positivelyaffect entry rates,but these effects are small.The results suggest the effects of government spending onproductivity improvements are,if any,relatively small.We contribute to the literature by documenting the differences in fiscal multipli
44、ers by the type of spending,whereas previous research has emphasized heterogeneity in economic conditions.Alonso(2017)and Duporet al.(2021)study the effect of fiscal policy on the composition of consumption.Our paper investigatesthe other side of the coin.Instead of exploring the heterogeneity in th
45、e consumption responses to shocks,we focus on the heterogeneity in the shocks themselves,i.e.,services and goods spending shocks.Weshow this heterogeneity is an important yet overlooked determinant of the spending multiplier.Boehm(2020)is a notable exception and studies the different output response
46、s to government investment andconsumption shocks.We classify spending into goods and services rather than consumption and investment.The importance of this classification is twofold.First,the US economy has shifted from a goods economy toa services economy.Studying the differences between these two
47、types of spending would help policymakersin designing effective fiscal interventions.Second,goods and services use different production technologies.Quantifying their effects separately helps to identify important determinants of the fiscal multiplier.Ourfindings provide complementary insights to th
48、ose already documented in the literature.Our paper also relates to the vast literature on the determinants of the fiscal multiplier.This literaturehas highlighted the role of the business cycle(Riera-Crichton et al.,2015;Su arez-Serrato and Wingender,2016;Buchheim et al.,2020),trade openness(Ilzetzk
49、i et al.,2013;Corbi et al.,2019),the exchange rate4regime(Born et al.,2013),population demographics(Basso and Rachedi,2021),households heterogene-ity(Hagedorn et al.,2019),labor market rigidity(Cole and Ohanian,2004;Gorodnichenko et al.,2012),automatic stabilizers(Dolls et al.,2012;Galeano et al.,20
50、21),public indebtedness(Ilzetzki et al.,2013),the degree of monetary policy accommodation(Woodford,2011),firm-size distribution(Juarros,2020),andthe direction of the intervention(Barnichon et al.,2020)in amplifying the response of economic activity topublic spending.This paper concentrates on the ch