《牛津經濟研究院：2024雇主投資對當地社區的社會經濟影響研究報告（英文版）（52頁）.pdf》由會員分享，可在線閱讀，更多相關《牛津經濟研究院：2024雇主投資對當地社區的社會經濟影響研究報告（英文版）（52頁）.pdf（52頁珍藏版）》請在三個皮匠報告上搜索。

1、THE SOCIOECONOMIC IMPACTS OF EMPLOYER INVESTMENTS ON LOCAL COMMUNITIES OCTOBER 2024 THE SOCIOECONOMIC IMPACTS OF EMPLOYER INVESTMENTS ON LOCAL COMMUNITIES Anek Soowannaphoom/S3The socioeconomic impacts of employer investments on local communitiesTABLE OF CONTENTSForeword 4Executive summary 61.Introd

2、uction 81.1 Background context 81.2 Motivation for and objectives of this study 82.How we assess the impact of large investments 102.1 A quick guide to our approach in three simplesteps 102.2 What impacts did wetest?102.3 How long did we track them for?112.4 Characteristics of large investments that

3、 we have modeled 123.Key findings and insights 163.1 Large investments spur significant growth in business formation 173.2 Evidence of a sustained decrease in the unemployment rate 183.3 Projects help to raise prosperity by boosting average earnings 193.4 Balance of evidence suggests an increase in

4、labor force participation 203.5 Clear substitution effects apparent within the healthcare insurance market 243.6 A reduction in the rate of violent crime 263.7 What does this all mean for Small-and medium-sized businesses?274.Final thoughts 28Appendix 1:summary of literature review findings 30Append

5、ix 2:empirical strategy 32Appendix 3:data assembly,formatting and cleaning 34Appendix 4:modeling results and sensitivity analysis 424The socioeconomic impacts of employer investments on local communitiesTHE SCOPE OF THIS RESEARCH This project investigates the socioeconomic impact of when a large emp

6、loyer invests in a local community by relocating or opening a new establishment.Socioeconomic impacts studied range from changes to wages and healthcare coverage to crime rates.Due to the datasets used and empirical approach chosen in this research,there was not an opportunity to explore the positiv

7、e impacts found in this report as it pertains to specific communities.Additional research should be done to explore these impacts and disaggregate the data,by gender,race,veterans status,and other key socioeconomic categories.Similarly,we could not quantify the positive impact small and mid-size bus

8、inesses(SMBs)expanding in a region might have.Evaluating if a number of SMBs opening in a given region has the same,more,or less impact than a large investment is an important area of future study.FOREWORD5The socioeconomic impacts of employer investments on local communitiesTong Stocker/S6The socio

9、economic impacts of employer investments on local communitiesEXECUTIVE SUMMARYOxford Economics has undertaken a groundbreaking research program that is designed to shed new light on the topic of the socioeconomic impact of large employers entering a local community.1 Clearly,there is no objective de

10、finition of what characterizes a“large investment.”Although the number of jobs created on-site represents a highly intuitive metric in this context,the precise threshold that should define an investment large is inherently subjective.Our choice of 1,250 jobs has been informed by scientific principle

11、s.These are described in more detail in the Appendix.WHY HAVE WE DONE THIS?When a large employer decides to invest in a local community by relocating or opening a new space,it is often accompanied by splash headlines about the number of jobs the project will create.These investments can take many fo

12、rms,including manufacturing plants,distribution centers,and medical facilities.While an immediate direct increase in employment is self-evident,this type of information does not help us understand to what extent these projects have lasting impacts on the lived experience of those in the community.It

13、 naturally begs the question:what happens next?In the United States,there is a heated debate around this question.Parties on both sides can hold strong opinions.Too often,unfortunately,these convictions are not based on a sober assessment of available quantitative evidence and,instead,are based on a

14、necdote and isolated examples.Moreover,as the well-known aphorism puts it,“the plural of anecdote is not data.”This study seeks to rectify this issue by filling gaps in the existing literature to provide an objective and data-driven assessment of what happens when a new large employer opens in a com

15、munity.In doing so,we hope that our research will strengthen the quality of discussion around new employer investments by providing data that reconnect this debate with communities and economic realities.HOW DID WE DO IT?In summary,our research methodology has three key steps as follows:1.Data colle

16、ction:first,we compiled dataon large investments in US counties andsocioeconomic outcomesstatistical measuresthat capture important aspects of communitylife.For this research,we have defined theformer as examples where an organizationcreated at least 1,250 additional jobs1 on-siteat a single establi

17、shment.2.Econometric modeling:next,we estimateda set of econometric models that capturehow large investments have affectedsocioeconomic outcomes.In simple terms,the models work by testing how the relativeperformance of the county changed beforeand after the investment compared to a setof benchmark c

18、ounties(that did not receiveinvestments).3.Results application:finally,we apply thefindings from our model to trace out theimpacts over time and to quantify the effect ofthe investments in more accessible terms.7The socioeconomic impacts of employer investments on local communitiesWHAT DID WE FIND?T

19、he findings from our research indicate that investments in the form of new large business openings bring substantial benefits to the host community.In summary,these investments increase shared prosperity and strengthen the social fabric in local communities.Moreover,addressing the“what happens next?

20、”research motivation,our results indicate that these are sustained benefits and,indeed,build over time.Specifically,after five years we find that counties which experience investments by large employers enjoy the following benefits compared to counties which did not receive such an investment:New bu

21、siness formation grew by 3.6%.We find very robust evidence that large investments stimulate business formation with an average increase in the number of business establishments of 3.6%compared to what otherwise would have occurred.In absolute terms,the typical impact of these large investments was t

22、o add nearly 100 new business establishments.Employee earnings rose by 1.4%on average,meaning greater prosperity for local workers.In absolute terms,the typical impact wasto boost workers wages by almost$635 annually,measured in todays prices.Unemployment rate is 0.4 percentage points lower.We find

23、that,on average,investments bring an immediate and sustained dropin the unemployment rate.In absoluteterms,the typical impact was to reduce the unemployment headcount by over 210 i.e.,just over 210 adults were in work who would otherwise have been unemployed.Participation rate is 2.1 percentage poin

24、tshigher.Additional dynamism and prosperityencourage workers to engage in the laborforce who would otherwise have beeninactive,equivalent to typical increase of1,560working-age adults.Violent crime rate is 8.1%lower.Potentiallylinked to the decline in the unemployment rate,we find an average reducti

25、on in the incidenceof violent crime of 8.1%five years after theinvestment.In absolute terms,this wasequivalent to an average drop of 140 recordedviolent crimes.Private insurance rate increases 1.3percentage points.We found evidence ofsubstitution effects within the healthcareinsurance market.On aver

26、age the share ofthe population on private insurance increased1.3 percentage points while there was acorresponding decrease in public insurance(e.g.,Medicaid)rates.In absolute terms,thetypical impact was to increase the numberof people with private healthcare by almost1,200.Given that we find no evid

27、ence ofany change in the overall rate of healthcarecoverage,it can be assumed that theseindividuals will be switching from publicinsurance programs such as Medicaid.8The socioeconomic impacts of employer investments on local communities1.INTRODUCTION1.1 BACKGROUND CONTEXT When a large employer decid

28、es to invest in a local community by expanding or opening a new space,it is often accompanied by splash headlines about the number of jobs the project will create on-site.While an immediate increase in employment is self-evident,this type of information does not help us understand to what extent the

29、se projects have lasting impacts on the lived experience of those in the community.It naturally begs the question:what happens next?In the United States,there is asometimes heateddebate around this question.Critics of large employers allege that these businesses do not apply a sufficiently holistic

30、model of stakeholder capitalism and squeeze out small-and medium-sized businesses(SMBs).As part of our research program,we have reviewed the relevant literature.A detailed summary of our findings can be found in Appendix One of this report.1.2 MOTIVATION FOR AND OBJECTIVES OF THIS STUDYIn this conte

31、xt,Oxford Economics sought out to assess the socioeconomic impact of investment projects on their local community.We designed the study so that it incorporated the following features:Comprehensive:given the priority on an industry inclusive assessment of large employers socioeconomic impact,it was e

32、ssential that our approach covered investments by businesses across the industrial spectrum of the private sector.Although our analysis contains a case study,this is used to add color to the reports narrative.On the other hand,our quantitative empirical strategy is designed to ensure that we capture

33、 investments by businesses in a wide range of economic sectors.Investments that move the dial:all examples of business investment,large and small,are a sign of economic progress and a driver of future prosperity.In this study,however,we have sought to evaluate the impact of investments that are quan

34、titatively“large”.This choice is motivated by seeking to be topicalthe impact of larger investments or investments by large employers is currently the source of more contentionand practical issuesidentifying the impact of smaller investments empirically is likely to be more challenging.Understanding

35、 the dynamics:where empirical evidence does exist it often provides a snap-shot assessment i.e.,at a single point in time.It is plausible,however,that the impact of large investment projects on local communities is not stable but instead evolves,influenced by a complex web of interacting factors.Our

36、 research method aims to cut through that complexity to help understand how these impacts change over time.Economic and social:much of the current literature has sought to establish the local impacts of large investments as measured by economic rather than social outcomes.In many respects,this is na

37、tural.The former can be expected to have more direct linkages with investment and data are typically more readily available.Our study seeks to address this gap by examining impacts through a broad socioeconomic lens.Overall,we hope that the study will raise the level of debate around new employer in

38、vestments and highlight to companies the importance of understanding their investments in terms of their broader potential community impacts.Too often,current discussions are framed around anecdote and as the well-known aphorism puts it,“the plural of anecdote is not data.”This study seeks to rectif

39、y this issue by applying an objective and data-driven method to answer these questions.9The socioeconomic impacts of employer investments on local communitiesdpa picture alliance/Alamy Stock Photo10The socioeconomic impacts of employer investments on local communities2.HOW WE ASSESS THE IMPACT OF LA

40、RGE INVESTMENTSIn this chapter,we provide a high-level description of our methodology.Our aim is to introduce our approach and explain key concepts as they relate to our findings using accessible and jargon-free language suited to a broad audience.The various appendices to this document contain a de

41、tailed technical description of our methodology for interested readers.2 Clearly,there is no objective definition of what characterizes a large investment.Although the number of jobs created on-site represents a highly intuitive metric in this context,the precise threshold that should define an inve

42、stment large is inherently subjective.Our choice of 1,250 jobs has been informed by scientific principles.These are described in more detail in the Appendix.3 Specifically,the share of working-age adults participating in the labor force was 63.3%in 2019 Q4 compared to 62.8%in September 2023.4 As Cas

43、e and Deaton note,the causal channel from unemployment to drug-taking is multi-factorial:“Our main argumentis that the deaths of despair reflect a long-term and slowly unfolding loss of a way of life for the white,less educated working class.Unemployment is a part of that story,but only a part.”Anne

44、 Case and Angus Deaton,2020.Deaths of Despair,Princeton University Press.Princeton,NJ,at p.146.2.1 A QUICK GUIDE TO OUR APPROACH IN THREE SIMPLESTEPSIn summary,our research methodology has three key steps as follows:1.Data collection:first,we compiled data on large investments in US counties and soc

45、ioeconomic outcomes.For this research,we have defined the former as examples where an organization created at least 1,250 additional jobs2 on-site at a single establishment.2.Econometric modeling:next,we estimated a set of econometric models that capture how large investments have affected socioecon

46、omic outcomes.In simple terms,the models work by testing how the relative performance of the county changed before and after the investment compared to a set of benchmark counties(that did not receive investments).3.Results application:finally,we apply the findings from our model to trace out the im

47、pacts over time and to quantify the effect of the investments in more accessible terms.2.2 WHAT IMPACTS DID WETEST?To develop a more holistic understanding of the effects of large investments,we have tested their impact on a range of outcome variables.Our selection of the precise indicators was infl

48、uenced by a desire to achieve a good balance of investigating impacts through both economic and social lenses and the constraints of available data.For our purpose,it was essential to be able to access data which monitored selected indicators over an extensive time horizon(ideally back to 1990)and a

49、cross a comprehensive set of US counties.The final selected set of outcome indicators for our models were as follows:Business dynamism:the number of establishments.A common contention of critics is that when firms undertake large investments,they force existing local firms out of business.On the oth

50、er hand,large investments may well spur new business formation by boosting the level of economic demand in the community.Labor force activity:the participation rate(%share of working-aged adults who are either in work or actively seeking work).A growing concern for policymakers since the Covid-19 pa

51、ndemic has been the associated slump in labor activity ratesindeed,at the time of writing,the US labor force participation rate remained 0.5 percentage points below its pre-pandemic peak.3 Joblessness:the unemployment rate(%share of the labor force that is not in work).Unemployment is of primary con

52、cern for local policymakers both because of its debilitating impact on affected individuals and families but also due to its association with various activities that can have wider negative social effects such as substance abuse.4 11The socioeconomic impacts of employer investments on local communit

53、ies Economic prosperity:average value of annual earnings.The previous two metrics help us to understand how these investments might affect the number of job opportunities created in the community.For policymakers,of wider interest,is whether these investments boost the prosperity of local residents.

54、Access to healthcare:healthcare insurance rates(%share of population covered by healthcare insuranceprivate and public).The extent and composition of healthcare insurance coverage is important.Although reforms enacted under President Obama have helped to significantly widen healthcare insurance cove

55、rage,individuals on employer-sponsored schemes typically benefit from superior service.In addition,enlarged public healthcare insurance rolls,such as for Medicaid,require significant state and federal expenditures.The incidence of criminal activity:number of violent crimes recorded per capita.Crime

56、is typically near top-of-mind for local policymakers with implications for how citizens feel about living in their community.5 For completeness,we have reported results over longer time periods in the Appendix of this document.2.3 HOW LONG DID WE TRACK THEM FOR?Data for these outcome variables that

57、we have tested are available over quite different time series in some cases.For four of the six indicators,we have data available for all years between 1990 and 2016.Data on labor force participation rates and healthcare coverage rates,however,are only available starting from 2005 and 2008,respectiv

58、ely.Moreover,we require at least four years of data on outcomes in a county before it receives an investment.This means that for healthcare coverage rates we are limited to investments that take place from 2012 to 2016 and for labor force participation from 2009 to 2016.These much more limited effec

59、tive sample sizes have implications for how we evaluate the path of the impact of these investments.For all six variables,the sample of investments that is investigated in years one to three after the investment is identical.As shown in Fig.1,for the four indicators with longer time series data,the

60、sample changes very modestly in years four to seven.Indeed,in year seven we can observe impacts for over 94%of the original sample,a share that drops to 47%for the participation rate variable and 0%for healthcare coverage.Balancing our interest in investigating impacts over longer horizonsgiven that

61、 they can be expected to accumulate over timewith the methodological issues that are created by these changes in samples size we opt for the following rules:5 For our analysis of the impact of these investments on the unemployment rate,the number of business establishments,average wages and the crim

62、e rate we report findings up to seven years after the initial investment.For our analysis of the impact of these investments on healthcare coverage and the participation rate we report findings up to five years after the initial investment.12The socioeconomic impacts of employer investments on local

63、 communities2.4 CHARACTERISTICS OF LARGE INVESTMENTS THAT WE HAVE MODELEDFor our modeling purposes,we are not able to include counties which received multiple investments in our model.Similarly,any counties which received a single investment before 1994 have also been excludedin these counties we do

64、 not have sufficient data on outcomes in the years leading up to the event to effectively conduct our“before-and-after”analysis.In total,we have identified 225 counties which received a single large investment between 1994 and 2016.The impact results in this report are estimated by comparing outcome

65、s in these counties to the aforementioned control group.In the remainder of this section,we briefly describe characteristics of theseinvestments.When did they take place?Fig.2 illustrates the number of investments that take place in each year across the sample at this threshold between 1994 and 2016

66、the effective modeling horizon.The year of the investment naturally affects the length of the period after which historic data are available to evaluate its impact.For example,for an investment that takes place in 1994,we have 25 years of pre-Covid-19 data and,at the other extreme,for one that takes

67、 place in 2016 there is just three years of data.In theory,we would expect that the impact of these investments is unlikely to remain constant in each post-treatment year and a key objective of our research was to understand the typical dynamic trajectory of the impact of these investments.In this l

68、ight,the fact that the investments in our dataset are relatively back-loaded,i.e.,they mainly take place in earlier years,is advantageous.Fig.1:Proportion of investments covered by dataset depending on post-event year and variable indicatorShare of single treatment events in sampleSource:Oxford Econ

69、omics12345670%10%20%30%40%50%60%70%80%90%100%Healthcare coverageParticipation rateOther variablesYears from investment,0=year of investment13The socioeconomic impacts of employer investments on local communitiesIn another respect,this back-loaded feature is potentially disadvantageous.It might incre

70、ase the risk that the results we estimatewhich reflect the average impact across all 225 investmentsare less representative(for better or worse)of the impacts that will be generated by recent and future investments.To assess this,we have conducted sensitivity analysis where we restrict our dataset t

71、o a more recent period(2005 2016).In general,our findings remain robust to this change indicating that our broad conclusions remain valid.Further discussion and the results from the sensitivity testing are presented in Appendix 4.Where did they take place?Fig.3 describes the geographic distribution

72、of large investments in our dataset using our preferred employment threshold.The counties that form part of our“single treatment”group and are,therefore,the subject of our modeling results,are shaded in orange.As displayed,they are widely disbursed across the country.As expected,counties that receiv

73、ed single or multiple investments are typically larger.For example,according to the latest Census Bureau estimates,in 2021 the(mean)average population of a county in the control group was 54,000 compared to an average of 125,000 for counties that received one investment and 590,000 for those who rec

74、eived two or more investments.Fig.2:Distribution of single treatment investments that created at least 1,250 jobs on-site:1994-2016Number of investments that create at least 1,250 jobs on-siteSource:NETS data,Oxford Economics analysis199419951996199719981999200020012002200320042005200620072008200920

75、1020112012201320142015201605101520253014The socioeconomic impacts of employer investments on local communitiesIn what industries do these investors operate in?Finally,Fig.4 breaks down the single treatment investments in our sample in terms of the economic sector of the investor.This highlights that

76、 our study cuts across a broad swathe of the economy with investments not concentrated in any particular industry.The most populous sector was financial and business services which accounted for 40 of the 225 single treatment cases.Approximately one-third of the investments were in manufacturing spr

77、ead across a very wide range of sub-sectors covering both consumer goods and industrial production.Fig.3:Overview of distribution of counties by classification statusFig.4:Breakdown of single treatment investments by economic sectorSource:NETS database,Oxford EconomicsNot in NETSControl groupSingle

78、treatmentMultiple treatment103236912204053229Agriculture and miningB2C manufacturingIndustrial manufacturingTransport equipmentTransport,communicationsand utilitiesWarehousing,storageand retailFinancial and businessservicesHospitality and leisureHealthcareMiscellaneousSource:NETS database,Oxford Eco

79、nomics analysisNumber of single treatment investments15The socioeconomic impacts of employer investments on local communitiesJohn Gress Media Inc/S16The socioeconomic impacts of employer investments on local communities3.KEY FINDINGS AND INSIGHTSThe headline findings from our analysis are that large

80、 investments spur local entrepreneurial activity and improve labor market conditions(higher wages,lower unemployment,and higher rates of participation),which in turn support wider social benefits(higher participation in private healthcare reducing the burden on the state and lower rates of violent c

81、rime):New business formation grew by 3.6%.We find very robust evidence that large investments stimulate business formation withan average increase inthe number of business establishments of 3.6%compared to what otherwise would have occurred.In absolute terms,the typical impact of these large investm

82、ents was to add nearly 100 new business establishments.Employee earnings rose by 1.4%on average,meaning greater prosperity for local workers.In absolute terms,the typical impact wasto boost workers wages by almost$635 annually,measured in todays prices.Unemployment rate is 0.4 percentage points lowe

83、r.We find that on average investments bring an immediate and sustained drop in the unemployment rate.In absolute terms,the typical impact was to reduce the unemployment headcount by over 210 i.e.,6 Specifically,we use the p-value of the average time treatment(ATT)coefficient over the truncated post-

84、treatment horizon.Full details on the statistical results from the model from each post-treatment year including the standard errors and p-values can be found in the Appendix to this document.just over 210 adults were in work who would otherwise have been unemployed.Participation rate is 2.1percenta

85、ge points higher.Additional dynamism andprosperity encourageworkers to engage in thelabor force who wouldotherwise have been inactive,equivalent to typical increaseof 1,560 working-age adults.Violent crime rate is 8.1%lower.Potentially linkedto the decline in theunemployment rate,wefind an average r

86、eductionin the incidence of violentcrime of 8.1%five years afterthe investment.In absoluteterms,this was equivalentto an average drop of 140recorded violent crimes.Private insurance rateincreases 1.3 percentagepoints.We found evidenceof substitution effects withinthe healthcare insurancemarket.On av

87、erage theshare of the population onprivate insurance increased 1.3percentage points while therewas a corresponding decreasein public insurance(e.g.,Medicaid)rates.In absoluteterms,the typical impact wasto increase the number ofpeople with private healthcareby almost 1,200.Given thatwe find no eviden

88、ce of anychange in the overall rate ofhealthcare coverage,it can beassumed that these individualswill be switching from publicinsurance programs suchasMedicaid.In the remainder of this chapter,we describe the major results in more detail of interest from our research and draw out their implications.

89、We present the findings for each of the six key outcome variables,introduced in chapter two,in turn.For each outcome,we describe our findings using the following three steps:1.As explained,the modelprovides us with a bestestimate of the average impactof these large investments ineach year.To illustr

90、ate howthese change over time weuse charts that plot a line ofbest fit through the preciseestimate for each year tocreate a smoothed impressionof these dynamic impacts.2.As with any econometricmodel,the estimated effectsin each year are subject tostatistical uncertainty.Toillustrate the extent of th

91、isuncertainty and to inform ourjudgement of the extent towhich our findings should betreated as statistically robustwe present and summarizeresults from standard statisticaltests which essentially describehow confident we are that theestimated effects are differentfrom zero.63.Finally,where we havei

92、dentified that the impact ofthe investment is statisticallyrobust,we apply the findingsof the model to historic datafor the 225 counties thatreceived a large investmentto quantify the typical impactthat counties have experiencedin absolute terms.17The socioeconomic impacts of employer investments on

93、 local communities3.1 LARGE INVESTMENTS SPUR SIGNIFICANT GROWTH IN BUSINESS FORMATION7 The approach we have used does not enable us to estimate the breakdown of these new establishments by size.8 See:Shai Bernstein,Emanuele Colonnelli,Xavier Giroud,and Benjamin Iverson.2019.Bankruptcy spillovers.Jou

94、rnal of Financial Economics 133(3):608-633,Efraim Benmelech,Nittai Bergman,Anna Milanez,Vladimir Mukharlyamov,The Agglomeration of Bankruptcy,The Review of Financial Studies,Volume 32,Issue 7,July 2019,Pages 25412586,and Daniel Shoag,Stan Veuger,2018.Shops and the City:Evidence on Local Externalitie

95、s and Local Government Policy from Big-Box Bankruptcies.The Review of Economics and Statistics;100(3):440453 We find that large investments have a large and positive impact on the level of business formation in the local county,as measured by an increase in the number of business establishments.Thes

96、e establishments could reflect new sites that are either formed by businesses that are already operating in the county or new firms entering the area for the first time.7 As shown in Fig.5,this impact steadily increases in the years following a large investment.For example,we estimate that after rec

97、eiving an investment the number of establishments in the county is 1.5%higher after three years,an impact that rises to 3.6%after five years and 4.5%after seven years.Our modeling,therefore,strongly rebuts the notion that the impact of large investments is,on net,to constrain existing local business

98、es and,to the contrary,that new entrepreneurial opportunities are created through businesses seeking to cluster together.This is broadly consistent with the existing empirical evidence in theliterature.8 On average,the counties which received one of these investments had 2,745 business establishment

99、s present five years after the event.This implies that the typical impact of these large investments was to add 96 new business establishments within five years.Fig.5:Change in the number of establishmentsFig.6:Confidence interval around estimated impacts on the number of establishments%change in nu

100、mber of business establishmentsSource:Oxford Economics estimates10234567Years from investment,0=year of investment0.0%0.5%1.0%1.5%2.0%2.5%3.0%3.5%4.0%4.5%5.0%Line of best fitModel estimates%change in number of establishmentsSource:Oxford Economics estimates10234567Years from investment,0=year of inv

101、estment0%1%2%3%4%5%6%7%18The socioeconomic impacts of employer investments on local communitiesAs with any econometric model these impacts are subject to uncertainty.Statistically,this can be gauged by a confidence interval which reflects the underlying variation in impacts that we observe.Fig.6 ill

102、ustrates these for each year.9 The length of the bars widens over time reflecting the precise size of the impact becomes more uncertain each year followingthe investment.Nonetheless,the fact that even the lower bound of these bars remains above zero in every post-investment year(apart from year zero

103、)indicates that our findings are highly robustindeed,the average annual impact was statistically significant at the 5%level.In summary,the evidence implies that we can be highly confident that one impact of large investments is to increase the number of business establishments in the county.9 All co

104、nfidence interval charts in this section are drawn for a 90%interval.This means that we can be 90%confident that the actual impact lies between the range presented.3.2 EVIDENCE OF A SUSTAINED DECREASE IN THE UNEMPLOYMENT RATEIn addition,we find clear evidence that large investments lead to sustained

105、 economic and social community benefits by lowering the rate of unemployment.As displayed in Fig.7,the impact of the investments on the unemployment rate seems to be relatively large and instantaneous and improves steadily before plateauing between years five and seven,implying that the gains are pe

106、rmanent.Specifically,we find that the unemployment rate falls by 0.30 percentage points in year two and is 0.38 percentage points lower in years five and seven.In our sample,the counties which received one of these investments had,on average,55,000 working-aged adults in the labor force five years a

107、fter the receipt of the investment.This implies that the typical impact of these large investments was to reduce the unemployment headcount by 210 i.e.,just over 210 adults were in work who would otherwise have been unemployed.Fig.7:Percentage point change in the unemployment rate%point change in un

108、employment rateSource:Oxford Economics estimates10234567Years from investment,0=year of investment-0.45%-0.40%-0.35%-0.30%-0.25%-0.20%-0.15%-0.10%-0.05%0.00%Line of best fitModel estimates19The socioeconomic impacts of employer investments on local communitiesAgain,model testing indicates that this

109、finding is statistically robust.Inverse to the results from the establishment model,the upper bound of the confidence interval remains below zero in every post-investment year indicating that we can be highly confident that the impact of these investments is to reduce the rate of unemployment in the

110、 host county.Across the post-treatment period,we found that the average annual impact was statistically significant at the 5%level.%point change in the unemployed rateSource:Oxford Economics estimates10234567Years from investment,0=year of investment-0.8%-0.6%-0.5%-0.4%-0.3%-0.2%-0.1%0.0%-0.6%Fig.8:

111、Confidence interval around the estimated impact on the unemployment rate3.3 PROJECTS HELP TO RAISE PROSPERITY BY BOOSTING AVERAGE EARNINGS10 Michael Greenstone,Richard Hornbeck,and Enrico Moretti.2010.“Identifying agglomeration spillovers:Evidence from winners and losers of large plant openings.”Jou

112、rnal of Political Economy,118(3):536598 These authors estimate an adjusted increase in wages of 2.7%in“winning”counties relative to losing after a Million Dollar Plant opening(at p.579).Contrary to some suggestions that large employers create low value jobs,our modeling indicates that there is an ov

113、erall uplift on average wages in the community.Fig.9 shows that the impact is found to increase steadily in the seven years following the investment.Specifically,we find that the effect of receiving a large investment is to raise average earnings by 0.9%after two years,with the impact increasing to

114、1.4%in year five and 1.7%in year seven.These results are broadly in line with findings of authors such as Greenstone,Hornbeck,and Moretti(2010).10Fig.9:Impact of large investments on average earnings in recipient county during years from investment%change in earningsSource:Oxford Economics estimates

115、10234567Years from investment,0=year of investment0.0%0.2%0.4%0.6%0.8%1.0%1.2%1.4%1.6%1.8%Line of best fitModel estimates20The socioeconomic impacts of employer investments on local communitiesOn average,in the counties which received one of these investments the average wage level five years after

116、the event was$45,200,measured in todays prices.This implies that the typical impact of these large investments was to boost workers wages by almost$635 annually within five years.As shown in Fig.10,the finding that large investments have a positive impact on average earnings in the host county is hi

117、ghly statistically robust.Again,our lower bound estimates are positive in every year and the average annual impact that we estimate through the seven-year period is statistically significant at the 5%level.3.4 BALANCE OF EVIDENCE SUGGESTS AN INCREASE IN LABOR FORCE PARTICIPATION Another important me

118、asure of labor market strength is the rate of labor force participationthe share of the population either working or looking for work.Reductions in the unemployment rate will understate the gross positive impact on employment,i.e.,the total number of jobs,if people who previously were not looking fo

119、r work rejoin the labor force and obtain jobs.The importance of this indicator has risen up policymakers agenda since the Covid-19 pandemic saw a sharp drop in the labor force participation rate which has yet to rebound.At the time of writing,the participation rate in September 2023 Q1 was 62.8%,0.5

120、 percentage points lower than the pre-pandemic peak rate of 63.3%.As displayed in Fig.11,our central estimates of the impact in the first five years following an investment are all positive and steadily increase with time.As noted in section two,due to data constraints,we curtail our analysis of the

121、 investment impact to the subsequent five-year period.After two years,our central estimate suggests that the investment led to a 1.5 percentage points increase in the participation rate,an impact that increased to 2.1 percentage points in year five.Such increases are notable in scale.Although we are

122、 not able to formally test or assess the underlying causes of this change,the described increase in earnings,which would improve incentives for workers,can be expected to contribute to this effect.Equally,the broader employment opportunities offered by the increase in the rate of new business format

123、ion might also incentivize adults to seek work.Finally,it is likely that these investments are associated with relatively large effects on inward migration,and it is plausible that this effect might partly reflect a change in the composition of the local population.%increase in average earningsSourc

124、e:Oxford Economics estimates10234567Years from investment,0=year of investment0.0%0.5%1.0%1.5%2.0%2.5%3.0%Fig.10:Confidence interval around the estimated impact on average earnings21The socioeconomic impacts of employer investments on local communities11 Note that this number differs from the one pr

125、esented when applying the results from the unemployment rate model because the sample of counties treated differed due to the constraints of available data.On average,in the counties which received one of these investments the working age population was 75,800.11 This implies that the typical impact

126、 of these large investments was to increase the number of working-age adults seeking work by 1,560 within five years.In general,however,the statistical performance of the model means that we should treat these findings with more caution compared to the impacts presented previously in this section.Fi

127、g.12 displays the confidence intervals around our central impact estimates in each year.The lower bound of these ranges goes below zero in all years except year three.Therefore,we can be statistically less confident that the impact of the investments on the participation rate is positive(compared to

128、 the other labor market effects on unemployment and earnings).%point change in labor force participation rateSource:Oxford Economics102345Years from investment,0=year of investment0.0%0.5%1.0%1.5%2.0%2.5%Line of best fitModel estimates%point change in the labor force participation rateSource:Oxford

129、Economics102345Years from investment,0=year of investment-2%-1%0%1%2%3%4%5%Fig.11:Impact of large investments on the labor force participation rate in recipient county up to five years from investmentFig.12:Confidence interval around the estimated impact on the labor force participation rate22The so

130、cioeconomic impacts of employer investments on local communitiesCASE STUDY:AMAZON FACILITY IN WILMINGTON,DELAWAREThis studys quantitative analysis establishes the socioeconomic impact of large employers across industries and over an extended time horizon.However,we also captured the practical impact

131、 of large employers at the individual establishment level through a case study analysis for an Amazon fulfillment center located in Wilmington,Delaware.This case study highlighted how the entry of a large employerin this case through an Amazon fulfillment centerrevived a shuttered work location and

132、created quality employment opportunities for local workers.12“Harvey Hanna&Associates and Dermody Properties Announce Plans for Redevelopment of Former General Motors Plant in Wilmington,Del.”November 8,2019.13 Maureen Milford,“GM closing Boxwood Road,last auto plant in Delaware,”The(Wilmington)News

133、 Journal,July 12,2009.https:/ 14 Ibid.15 Harvey Hanna&Associates,“Harvey Hanna&Associates Submits Exploratory Plan for Redevelopment of Former General Motors Plant in Wilmington,Delaware,”April 16,2018.https:/ See https:/ Jacob Owens,“Delaware OKs$4.5M Amazon grant for Boxwood,”Delaware Business Tim

134、es,February 24,2020.18 Brandon Holveck,“With 3,000 employed at Boxwood mega-warehouse,Amazons Delaware presence continues to grow,”Delaware News Journal,4 March 2022 https:/ Brandon Holveck,”Amazon starts hiring seasonal employees in Delaware but only at certain facilities,”Delaware News Journal,19

135、October 2021 https:/ Brandon Holveck,“With 3,000 employed at Boxwood mega-warehouse,Amazons Delaware presence continues to grow,”Delaware News Journal,4 March 2022 https:/ A STORIED LOCATIONIn 2019,Dermody Properties acquired 88 acres of the 142-acre property on Boxwood Road to develop what would be

136、come the Amazon MTN1 fulfillment center.12 The previous GM assembly plant,opened in 1946,was integral to the community but was shuttered in 2009 during the financial crisis as GM emerged from bankruptcy.13 The plant was a“$182 million-a-year economic engine”according to Delaware Economic Development

137、 Office estimates14 which employed 6,000 people at its height.15 A nearby Chrysler plant closed around the same time.The Delaware Prosperity Partnership(DPP)is the nonprofit state economic development agency leading state efforts to grow local business and works with private employers on investment

138、opportunities and around funding opportunities.16 According to DPP President and CEO,Kurt Foreman,DPP helped Dermody navigate the permitting application process,as well as the grant application with the Delaware Economic Development Authority.Specifically,Amazon received a$4.5 million grant in Febru

139、ary 2020,including$3 million tied to“hiring 1,000 new full-time positions over the next three years and$1.5 million towards the fit-out of the nearly 3.7 million-square-foot facility being built by Dermody Properties.”17 Amazon and Dermody developed the facility into a major nearly 4 million square

140、feet robotics fulfillment center over five levels which opened in September 2021.THE FACILITY REPRESENTS A SIGNIFICANT SOURCE OF LABOR DEMANDAmazon has exceeded the grant award hiring targets and employs 3,000 people,18 with higher numbers during peak seasons.In addition,Amazon offers workers of all

141、 backgrounds and with little to no formal post-secondary training relatively high wages and comprehensive benefits,including full medical and dental insurance,401k with match,and paid leave.In its launch year,the average starting wage at MTN1reached$18/hour19 with signing bonuses“up to$3,000”.20 An

142、individual making$18/hour on a full-time basis earns approximately the average per capita income in Delaware.23The socioeconomic impacts of employer investments on local communitiesHIRING FROM THE LOCAL COMMUNITYCharles A.Madden,of the Delaware Prosperity Partnership,noted that Amazon made broad out

143、reach to communities surrounding the Boxwood Road location.In partnership with local elected officials,they helped organize job fairs highlighting Amazon roles at a high school and community centers.In addition,open houses were arranged,during which there was on-site support for potential applicants

144、 who might not otherwise have access to computers for completing online applications.Specifically,Amazon came on site to Career Team Delawares facility and provided coaching to people on how to complete the forms.Career Team Delaware is an organization that provides Employment and Training for SNAP

145、and TANF recipients in the State of Delaware.Mr.Madden explained also how DPP helped make connections with people in the community.This involved bringing together community-based leaders and facilitating Amazon giving presentations on roles.21 Ibid UPSKILLING OF LOCAL WORKFORCEAmazon has expressed a

146、$1.2 billion commitment to investing in human capital in the form of workforce upskilling including via its Career Choice program.Amazon pays tuition for hourly employees to attend programs through their education partners,which include national providers as well as the University of Delaware,Wilmin

147、gton University,and Delaware Technical Community College.21 We understand that this benefit can help to reduce turnover,even as workers prepare for careers outside of Amazon.We received from Amazon a testimonial from one of their employees,Emily,who after starting at the Boxwood facility and learnin

148、g about the Career Choice program made the decision to pursue an associate degree in cybersecurity and IT at Wilmington University(online).Emily shared that she has had“past employers offer tuition reimbursement,but Amazon was the first to make it clear,encouraging,and easy to sign-up.”She recently

149、enrolled in her fourth course and was promoted to a Seasonal IT Equipment Coordinator.Amazon Delaware Khairil Junos|D24The socioeconomic impacts of employer investments on local communities3.5 CLEAR SUBSTITUTION EFFECTS APPARENT WITHIN THE HEALTHCARE INSURANCE MARKET 22 See Medicaid.gov,“July 2023 M

150、edicaid&CHIP Enrollment Data Highlights,”available at https:/www.medicaid.gov/medicaid/program-information/medicaid-and-chip-enrollment-data/report-highlights/index.html,accessed November 2023.Reforms enacted during the Obama administration led to a dramatic increase in the overall rate of healthcar

151、e insurance,with many of the previously uncovered households now benefiting from publicly sponsored schemes,notably Medicare and Medicaid.On the other hand,private healthcare insurance remains largely accessed through employer-sponsored schemes.Medicaid itself was expanded as part of the ACA,with co

152、verage extending to low-income adults up to 138%of the Federal Poverty Level.As of July 2023,Medicaid/CHIP enrollment was at 92 million.22 The evidence presented so far in this chapter demonstrates that these large investments have positive impacts on the labor market which can be expected to result

153、 in an increase in the employment rate i.e.,the share of working-aged adults in the county who are in work.Taken together,we would expect,therefore,that the impact of these investments should be to create a substitution effect with a shift from public toward private health insurance coverage in the

154、recipient county if the jobs created are high quality and come with associated health benefits.As displayed in Fig.13,this is indeed what our modeling suggests.The impact on the overall rate of healthcare insurance is negligible with the grey shaded line remaining close to zero through the five-year

155、 evaluation period.This,however,masks a notable switch between the share of the population on private(dark blue line)rather than public(light blue line)schemes.Specifically,after two years private healthcare coverage rates are 1.0 percentage points above where they would have been and around 1.3 per

156、centage points up after five years.Fig.13:Impact of large investments on healthcare coverage rate in recipient county up to five years from investment%point change in healthcare coverage by typeSource:Oxford Economics estimates102345Years from investment,0=year of investment-2.5%-2.0%0.0%-1.5%0.5%1.

157、0%1.5%2.0%Private-line of best fitPrivate-Model estimatesPublic-line of best fitPublic-Model estimatesOverall-line of best fitOverall-Model estimates-0.5%-1.0%25The socioeconomic impacts of employer investments on local communitiesIn general,these findings invert some popular perceptions of large em

158、ployers as taking advantage of government benefits like Medicaid;instead,our analysis suggests that exactly the opposite is the case.While large employers might hire a number of workers who join the business while receiving public benefits,their investments ultimately increase the proportion of peop

159、le receiving private healthcare coverage and lower the proportion of people receiving publicly provided healthcare coverage,such as Medicaid.These large investments on average may reduce the fiscal burden related to public healthcare coverage for impacted localities.On average,in the recipient count

160、ies the number of people insured by private healthcare five years after the investment was 91,100.This implies that the typical impact of these large investments was to increase the number of people with private healthcare by almost 1,200 within five years.Given that we find no evidence of any chang

161、e in the overall rate of healthcare coverage,it can be assumed that these individuals will be switching from public insurance programs such as Medicaid.Results from statistical testing for the private healthcare coverage rate impact estimates are presented in Fig.14.Similar to the results for the la

162、bor force participation model,our estimated impact range frequently dips below zero,albeit typically by a relatively small margin.Stepping back,the modeling results clearly indicate that our findings for the impact of large investments on both the labor force participation rate and healthcare covera

163、ge rates are subject to a higher degree of uncertainty compared to the results for business establishment formation,the unemployment rate,and earnings.We believe that this is likely to be primarily driven by the fact that we have a smaller effective sample of investments on which to test the impacts

164、 in these two models which naturally leads to greater statistical uncertainty.Overall,therefore,we consider that the balance of evidence strongly points towards large investments having a positive impact on both the labor force participation and private healthcare coverage rates.Fig.14:Confidence in

165、tervals around the estimated impact on the private healthcare coverage rate%point change in the private healthcare coverage rateSource:Oxford Economics estimates102345Years from investment,0=year of investment-1.5%-1.0%0%1.5%2.0%2.5%3.0%3.5%0.5%1.0%-0.5%26The socioeconomic impacts of employer invest

166、ments on local communities3.6 A REDUCTION IN THE RATE OF VIOLENT CRIME23 United Nations Office on Drugs and Crime,“Monitoring the Impact of Economic Crisis on Crime”,2012The findings from our violent crime model are displayed in Fig.15.As shown,the estimated impact is a persistent reduction in the r

167、ate of violent crime.There is a strong immediate impact which gradually advances before plateauing from year five onwards.The shape of this trajectory is broadly similar to the estimated impact on the unemployment rate,which is intuitive given that this,together with income inequality and deprivatio

168、n,are economic indicators that have most commonly been causally linked to crime rates.23Specifically,our smoothed central estimates point to a 9.1%decrease in the violent crime rate by year two,an 8.1%reduction in year five,and a 7.8%reduction in year seven.On average,in the recipient counties there

169、 were 1,564 violent crimes recorded in the single treatment counties five years after the investment.This implies that the typical impact of these large investments was to reduce the number of violent crimes by 137 within five years.Results from statistical testing are displayed in Fig.16.These show

170、 that in three-quarters of the post-treatment years we can be highly statistically confident that the investments impact was to reduce the rate of violent crime.Compared to the estimated impacts in other models,the effects are a bit more erratic,but this is not surprising given that actual data on c

171、rime rates tend to be relatively more volatile from year to year.Such volatility is a reason why detecting the true drivers of criminal activity has proved to be challenging in the social science literature.Overall,we think that the evidence strongly corroborates the view that the impact of large in

172、vestments is to result in a reduction in the rate of violent crime in the host county.Although the precise magnitude of this decrease is relatively uncertain,the balance of evidence points to it being sizeable.Fig.15:Impact of large investments on the rate of violent crime in recipient county up to

173、seven years from investmentFig.16:Confidence intervals around the estimated impact on the violent crime rate.%point change in the violent crime rateSource:Oxford Economics estimates10234567Years from investment,0=year of investment-14%-12%-10%-8%-6%-4%-2%0%Line of best fitModel estimates%change in t

174、he violent crime rateSource:Oxford Economics10234567Years from investment,0=year of investment-25%-20%-15%-10%-5%0%5%27The socioeconomic impacts of employer investments on local communities3.7 WHAT DOES THIS ALL MEAN FOR SMALL-AND MEDIUM-SIZED BUSINESSES?24 Technically,SBA size thresholds depend on

175、industry,and are based on average annual receipts or employee numbers.https:/www.sba.gov/document/support-table-size-standards 25 To the extent that a large investment can help to catalyze the production of specialized non-tradable inputs,there are agglomeration externalities that new entrants might

176、 benefit from with input sharing and increasing returns to scale in those inputs.See Moretti,E.,“Local Labor Markets,”in Handbook of Labor Economics,Elsevier,2011 at p.1290 describing“thick market(s)for intermediate inputs”.26 Moretti,E.,“Local Labor Markets,”in Handbook of Labor Economics,Elsevier,

177、201127 Goolsbee,A.,and A.B Krueger,“A Retrospective Look at Rescuing and Restructuring General Motors and Chrysler,”Journal of Economic Perspectives,Volume 29,Number 2,Spring 2015,Pages 324As highlighted,our findings describe the impacts of large investments on county level socioeconomic outcomes,wi

178、th the former defined as those which met a threshold of 1,250 new jobs created on-site.Given the prevailing statistical definition for small-and medium-sized businesses(SMBs)generally,firms with fewer than 500 employeesit can be safely assumed that none of these investments were by SMBs(at least at

179、the point of completion).24 Nonetheless,we believe that our empirical findings are of considerable relevance to this community.For example,our results indicate that these investments can spur business formation,finding a lift of 3.6%in new businesses,five years after the investment.In our data,that

180、amounts to around 100 new businesses in the average county over five years,many of which are likely to be SMBs.There are many potential mechanisms which might explain this impact.For example,the new investment might develop local supply chains to support that investment.25 Additionally,the jobs crea

181、ted and increased spending power from higher local wages might stimulate demand,and supply to meet that demand.An expanded pool of skilled workers(increased“labor market thickness”26)might make the county more attractive for investment.Furthermore,large new investments might be a signal to other com

182、panies deciding where to locate new production that the county is a suitable location.Our overall findings were specific to the effects of large investments,meeting a threshold of 1,250 new jobs created.This is not,however,to suggest that similar relative impacts would not be experienced with smalle

183、r investments from small-and medium-sized businesses.The impact of ten 125-job investments may be as large or greater than one 1,250 jobinvestment.In practice small and large firms operate in interdependent networks.Large investments might be a spur for demand;small businesses might provide speciali

184、zed non-tradable services that are inputs into a larger firms production process or offer services and amenities to employees of those large firms that make counties attractive locations for investment.These linkages explain why,for example,both the Bush and Obama administrations decided to rescue G

185、M and Chrysler following the financial crisis,as there were“widespread spillovers into supplier industries and auto dealerships as well as knock-on macroeconomic effects.”27 These agglomeration externalities help to explain also the clustering of companies and suppliers in regions like Silicon Valle

186、y,the Puget Sound,and the Research Triangle.These examples help to illustrate the interlinked nature of local economies.Location-based policies must recognize that local labor forces are somewhat mobile,and that small-and medium-sized businesses are integral to these economic networks.This project w

187、as focused on large investments as a starting point but teasing out the role of SMBs is an important area for future research.28The socioeconomic impacts of employer investments on local communities4.FINAL THOUGHTSWe believe that our findings are notable for both their breadth(the number of outcome

188、variables assessed)and their depth(the large number of large investments evaluated).This is not to diminish the value of existing studies,many of which apply a more focused lens but just to highlight how we have added value.The results from our modeling demonstrate that large-employer investments br

189、ing substantial benefits to the host community.These benefits go beyond the initial jobs impact and appear to stem initially from the injection of economic dynamism spurred by the projects,including higher rates of entrepreneurialism,improvements in labor market engagement and worker prosperity,and

190、wider social effects such as shifts to private healthcare insurance and a lower rate of violent crime.Our Amazon case study illustrated some of the mechanisms that might be at work.Investments can revive struggling or abandoned economic zones in communities.They can also spur demand for labor,build

191、workforce skills,and increase labor earnings.The specific pathways by which large employers benefit communities will vary,but in aggregate this analysis tells a consistent story of positive impact.Although our modeling approach does not allow us to formally test interrelationships,it can be expected

192、 that the various impacts that we have identified influence each other.For example,higher rates of labor market activity are almost certain to be a primary driver of shifts in healthcare insurance from public to private schemes and the associated reduced burden on the State.Similarly,unemployment ha

193、s been identified previously as one of the key drivers for individuals resorting to criminal activity and so the sustained decrease in joblessness can be expected to have contributed to lower rates of violent crime.Ultimately,we hope that our research efforts can be a spur to further quantitative re

194、search that can support efforts to raise the tone and level of the current debate.In particular,we think it is important to understand better how the positive impacts found in this report pertain to specific communities.Specifically,additional research that explores how these impacts are distributed

195、 within host communities disaggregating between households and individuals depending on their gender,ethnicity,veteran status etc.would be of great value.Indeed,we would love to play an active role in this supplementary research effort.Because for the residents of these local communities,whose voice

196、 is sometimes lost,it seems clear that attracting more major investment projects can yield tangible gains to their everyday lived experience.29The socioeconomic impacts of employer investments on local communitiesPetinov Sergey Mihilovich/S30The socioeconomic impacts of employer investments on local

197、 communities30The socioeconomic impacts of employer investments on local communitiesAPPENDIX 1:SUMMARY OF LITERATURE REVIEW FINDINGSVarious strands of economic literature touch on the impacts of large employers.This includes literature on investment and agglomeration externalities,28 the impacts of

198、the loss of anchor employers,29 the entry and expansion of companies like Walmart across the country,30 and the literature on state and local economic incentives.31 While this literature is growing and important,it has not always featured in the policy debate.28 Greenstone,Michael,Richard Hornbeck,a

199、nd Enrico Moretti.2010.“Identifying agglomeration spillovers:Evidence from winners and losers of large plant openings.”Journal of Political Economy,118(3):53659829 Shai Bernstein,Emanuele Colonnelli,Xavier Giroud,and Benjamin Iverson.2019.“Bankruptcy spillovers.”Journal of Financial Economics 133(3)

200、:608-633.30 Alessandra Bonanno,A.,&Stephan Goetz,(2012).“Walmart and Local Economic Development:A Survey.”Economic Development Quarterly,26(4),285297.https:/doi.org/10.1177/0891242412456738.The authors survey a number of papers finding results in both directions”for outcomes including:(a)retail(and

201、nonretail)businesses,(b)retail workers,wages,and types of jobs;and(c)producer and consumer welfare;as well as poverty rates,social capital,food insecurity,policy effectiveness,and obesity.They conclude that we are“far from understanding”long-run impacts.Baskers(2007)review of the impact of the entry

202、 of big box retailers like Walmart entering a local market finds a small increase in net job creation(around 400 new jobs are created by the median new Walmart store)with job losses at smaller retailers who have lower profitability through lower prices for consumers in the local market.Basker,E.(200

203、7).“The Causes and Consequences of Wal-Marts Growth.”Journal of Economic Perspectives 21(3),17719831 Cailin Slattery and Owen Zidar.2020.“Evaluating State and Local Business Incentives.”Journal of Economic Perspectives,34(2):90-118.Slattery and Zidar(2020)find an average discretionary subsidy of$178

204、 million for 1,500 promised jobs($120,000 per job).See also Slattery(2022)which provides a detailed overview of the literature on subsidy competition and who concludes that overall“state and local governments would be better off in the absence of subsidy competition.”Slattery,C.“Bidding for Firms:Su

205、bsidy Competition in the U.S.,”Working Paper,31 December 202232 Michael Greenstone,Richard Hornbeck,and Enrico Moretti.2010.“Identifying agglomeration spillovers:Evidence from winners and losers of large plant openings.”Journal of Political Economy,118(3):536598 They find an adjusted increase in wag

206、es of 2.7%in“winning”counties relative to losing after the plant opening(at p.579).These authors find a 12 percent increases in total factor productivity(TFP)among incumbent plants in“winning”counties that attracted a large manufacturing plant relative to incumbent plants in runner-up counties five

207、years after opening.This productivity increase they note“implies an additional$430 million in annual county manufacturing output 5 years after the million-dollar plant opening”(p.589).33 Shai Bernstein,Emanuele Colonnelli,Xavier Giroud,and Benjamin Iverson.2019.“Bankruptcy spillovers.”Journal of Fin

208、ancial Economics 133(3):608-633.“When the bankrupt establishment is large relative to the census block,we estimate that liquidation leads to a reduction of 1.851.97 jobs in the block per job at the bankrupt establishment”(p.610).34 Efraim Benmelech,Nittai Bergman,Anna Milanez,Vladimir Mukharlyamov,“

209、The Agglomeration of Bankruptcy,”The Review of Financial Studies,Volume 32,Issue 7,July 2019,Pages 25412586“Stores located in proximity to stores of national chains that are liquidated are more likely to close themselves”(p.2544)and“being located at the same address as a liquidating retail chain sto

210、re increases the probability of closure by 0.36 percentage points,or 5.9%of the sample mean.”35 Daniel Shoag,Stan Veuger,2018.“Shops and the City:Evidence on Local Externalities and Local Government Policy from Big-Box Bankruptcies.”The Review of Economics and Statistics;100(3):440453 at p.44336 Vik

211、ram Pathania and Serguei Netessine,“The Impact of Amazon Facilities on Local Economies”(October 26,2022).Available at SSRN:https:/ examine county level outcomes and find an employment to population ratio increase of 0.7-1.0%points(and a county-level jobs multiplier of around 1.9)from the opening of

212、a new Amazon distribution facility.This literature finds some evidence of spillovers from large investments including local employment increases.Greenstone,Hornbeck,and Moretti(2010)find increased labor costs(and higher productivity)among incumbent plants in counties that attracted a large manufactu

213、ring plant relative to incumbent plants in counties which narrowly missed out on these investments.32 Bernstein et al(2019)examine the impacts of bankruptcy liquidations and find employment declines in the neighborhood of the establishment particularly in nontradable and service sectors(restaurants

214、and retail)which they note is consistent with lost knowledge spillovers and reduced local traffic of consumers.33 Benmelech et al(2019)similarly show how a retail chain liquidation spills over onto other stores by reducing the attractiveness of retail centers for remaining stores.34 Shoag and Veuger

215、(2018)find spillover bankruptcies in the neighborhoods of a chainwide bankruptcy coupled with relative employment decreases.35 And beyond employment gains(or losses in the case of firm shuttering),there is also some evidence of increases in labor force participation from new establishments.3631The s

216、ocioeconomic impacts of employer investments on local communitiesBeyond labor market impacts,Hu,Tsang,and Wan(2021)look at the impact of corporate headquarter relocation events on the housing market and find that a move in headquarters is associated with a zip code level increase in housing price gr

217、owth.37 Draca and Machin(2015)38 review the literature on the relationship between crime and economic incentives.37 Maggie Hu,Desmond Tsang,and Wayne Xinwei Wan,“Corporate Relocation and Housing Market Spillovers”(May 11,2021).Available at SSRN:https:/ Mirko Draca,and Stephen Machin,2015,“Crime and

218、Economic Incentives”,Annu.Rev.Econ.2015.7:38940839 See also Raphael S,Winter-Ebmer R.2001.“Identifying the effect of unemployment on crime.”J.Law Econ.44:25984;and Lin M-J.2008.“Does unemployment increase crime?Evidence from US data 19742000.”J.Hum.Resour.43:41336.However,Chalfin and McCrary find th

219、at the impact of unemployment on property crime does not seem to be uniform over time.Aaron Chalfin,and Justin McCrary.2017.“Criminal Deterrence:A Review of the Literature.”Journal of Economic Literature,55(1):5-48They note wide ranging work on the relationship between unemployment and wages on youn

220、g male criminal behavior and find significant effects of unemployment on property and violent crimes.3932The socioeconomic impacts of employer investments on local communitiesAPPENDIX 2:EMPIRICAL STRATEGYThe aim of this project is to identify the investment impact from large companies on socioeconom

221、ic and macroeconomic indicators.To meet this requirement,we use a leading practice estimator designed to estimate treatment effects and causal impacts.MOTIVATIONTo be robust and fit for purpose,our model will need to account for the following features:Variations in treatment timethe time of arrival

222、of a large employer varies across counties The same county is treated multiple timesthe same county can experience multiple investments either in the same year or across different years Potential lag in treatment impact on outcome variablethere is a lag between the arrival of a large employer and th

223、e hypothesized impact of such arrival Potential dynamics in treatment effects(i.e.,treatment effects changing over time)it is plausible that the treatment effect does not stay constant over time but evolves40 Brantly Callaway and Pedro Santanna(2020)“Difference-in-Differences with Multiple Time Peri

224、ods”,https:/ Andrew Goodman-Bacon et al(2021)“Difference-in-Differences with a Continuous Treatment”,https:/ Potential dynamics in outcome variablespast investments might influence the likelihood of future investments Historically,this type of analysis has relied on the standard difference-in-differ

225、ence(DiD)model estimated using a two-way fixed effects(TWFE)method.Recent research from Callaway and SantAnna(2020)40 and Goodman-Bacon(2021)41 among others indicate that estimates from the TWFE method tend to suffer from severe bias especially in instances where there are variations in treatment ti

226、me.THE CALLAWAY AND SANTANNA ESTIMATORWe follow Callaway and SantAnna(2020)and consider identification,estimation,and inference procedures for treatment effect parameters using DiD with(i)multiple time periods,(ii)variation in treatment timing,and(iii)when the“parallel trends assumption”holds potent

227、ially only after conditioning on observed covariates.Note that this approach allows causal effect parameters to be identified in staggered DiD setups,even if differences in observed characteristics create non-parallel outcome dynamics between groups(see Callaway and SantAnna,2020).Our proposed estim

228、ator has at least two advantages over a TWFE regression.First,it is not biased by time varying treatment effects because it only uses untreated comparison groups.Second,it aggregates the cohort-specific treatment effect parameters by the share of treated units,while TWFE weights subgroup parameters

229、by treatment variances as well.See Goodman-Bacon(2021)and Sun and Abraham(2020)for a fuller discussion.IDENTIFYING A BASELINE TREND IN SINGLE TREATED COUNTIESWe require treated counties to have at least four years of data prior to(and post)treatment for the Callaway&SantAnna estimator to detect a pr

230、e-treatment trend.Without enough pre-treatment data,it would not be possible to estimate the causal impact of treatment on economic indicators.Therefore,single treated counties treated prior to 1994,or after 2016,are removed from the sample.This led to 323 counties which received at least one enterp

231、rise investment between 1991-1993 being discounted.The socioeconomic impacts of employer investments on local communitiesBilanol/S34The socioeconomic impacts of employer investments on local communitiesAPPENDIX 3:DATA ASSEMBLY,FORMATTING AND CLEANINGDATA ASSEMBLY,FORMATING,AND CLEANING42 A recent in

232、teresting application of this dataset is contained in Esteban Rossi-Hansberg,Pierre Daniel Sarte,and Nicholas Trachter,2021.“Diverging trends in national and local concentration.”NBER Macroecon.Annu.35:11550.See also Keith Barnatchez,Crane D,Leland,and Ryan A.Decker(2017).“An Assessment of the Natio

233、nal Establishment Time Series(NETS)Database,”Finance and Economics Discussion Series 2017-110.Washington:Board of Governors of the Federal Reserve System,https:/doi.org/10.17016/FEDS.2017.110.This paper shows the strong correlation between the Census Bureaus County Business Patterns and NETS firm-le

234、vel data.43 It is worth noting that,our sample included examples where that expansion occurred at multiple locations within a single county in that time period for a firm.For example,in the case where two establishments with the same HQ identifier added 125 employees each in a year in the same count

235、y were captured in our sample.To estimate our models,we collected data that would enable us to identify where(in which county)and when(in which year)“large investments”have taken place in the United States and match that to equivalently structured datasets that track various macroeconomic socioecono

236、mic outcomes.In this section,we describe the sources that we have used starting with investment data.Identifying the arrival of largeemployersThere is no straightforward definition of what should be characterized as a“large investment.”In this study,we have sought to identify historic examples based

237、 on the number of additional jobs that were created on-site.By using the number of extra jobs,we do not restrict ourselves to new facilities that are built from scratch but also allow for sufficiently large expansions to existing sites.The source we used was the National Establishment Time Series(NE

238、TS)database.42 The NETS database is created from Dun&Bradstreet firm level data by Walls&Associates and represents a comprehensive annual census of establishments across the United States.We used the data to identify historic examples where employers had added at least 250 new jobs at a single estab

239、lishment over any three-year time period.43 Clearly the employment threshold that is judged to represent a“large”investment is subjective.Fig.17:Counties in the United States covered in the restricted NETS database sampleSource:NETS database,Oxford EconomicsNot in NETSControl group,single treatment,

240、and other counties35The socioeconomic impacts of employer investments on local communitiesJim West/Alamy Stock Photo36The socioeconomic impacts of employer investments on local communitiesBy acquiring data on the number of new jobs at each establishment,we left ourselves flexibility to use and test

241、effects at higher thresholds.A discussion of how we determined an appropriate rule to define a large investment given the properties of our dataset.Although the underlying NETS dataset is highly comprehensive of historic economic activity in the United States,the restricted sample that was provided

242、to us means that certain counties were excluded from our analysis from the outset.In total,of the just over 3,000 US counties,approximately one-third were not included in the NETS data(because an employer had not added 250 jobs at an establishment in a single year).The geographic distribution of cou

243、nties in our restricted sample is displayed in Fig.17.This highlights that counties which do not feature in the dataset are predominantly clustered in the Midwest,West,and Southeast regions on the country.Consistent with this geographic pattern,counties that were excluded are typically very small.Th

244、e(mean)average population of this group of counties in 2021 was just 11,172,more than 13 times smaller than the average size of those counties that did feature(Fig.18).Finally,since our focus in this project was to evaluate the effects of private sector investments,we have excluded government-owned

245、enterprises from our event list.These were identified based on a variety of factors including:the company name;the name of the chief executive officer(CEO)(frequently a mayor,Governor,or the President might be listed);and a manual review of establishments with public administration standard industri

246、al classification(SIC)codes that we confirmed to be government owned.Defining the large investment thresholdAs mentioned,there is no objective definition of what should be characterized as a“large investment”.We believe that the number of new jobs created on siteas defined by the NETS databaseprovid

247、es an intuitive starting point but it still leaves a question of the appropriate threshold that should be used for the purpose of classification.All else equal,it would be preferable to use a lower employment threshold.The benefit of a lower threshold is that it increases the generalizability of our

248、 findings because more counties historically will have experienced investments of this nature.As we go on to explain,however,this comes with methodological disadvantages.In this section,we describe this trade-off in more detail and explain our rationale for our preferred rule.Fig.18:Average populati

249、on of counties included and excluded from the NETS database in 2021Population of county(thousands)Source:NETS database,Oxford Economics analysisMean020100120140160180608040MedianIncluded in our sampleMissing from our sample152,69811,17245,9868,44237The socioeconomic impacts of employer investments o

250、n local communitiesIn our modeling framework,we are constrained to only using counties which experience a single large investment during the modeling horizon.Those counties which experience two or more investments of sufficient size during this period,unfortunately,need to be excluded from our analy

251、sis.Fig.19 describes how the distribution of counties in our sample varies as the employment threshold for a qualifying large investment varies using the following four categories:Control:are counties which do not receive any investments that creates at least the specified number of jobs on-site dur

252、ing the modelling horizonin economics jargon this is our control group.44 Numbers may not sum to 100%due to rounding.Single treatment:are counties which receive one investment that creates at least the specified number of jobs on-site during the modelling horizon.Multiple treatment:are counties whic

253、h receive multiple investments that creates at least the specified number of jobs on-site during the modelling horizon.Treated before 1994:are counties which received at least one investment that creates at least the specified number of jobs on-site between 1990 and 1993.For these counties,there was

254、 insufficient data available to observe outcomes before the investment.As described,counties categorized in both the“multiple treatment”and“treated before 1994”groups had to be effectively discarded from our model.As shown in Fig.19,the share of counties in both these groups steadily falls as the em

255、ployment threshold increases in size.For example,at a threshold of 250,around 88%of our sample would have been discarded,a share that drops to 63%at an employment threshold of 500 and again to 46%at 750 etc.In this sense,it was methodologically advantageous to select a threshold that was relatively

256、large because it increased our effective sample size.Fig.19:Distribution of counties depending on the employment threshold for a large investment44Share of US counties in sampleSource:NETS database,Oxford Economics5002501000125015001750New jobs threshold0%20%60%30%80%70%90%100%ControlMultiple treatm

257、ent50%40%Single treatmentTreated before 1994750200010%2%9%25%63%23%14%18%45%39%15%10%36%53%11%6%30%61%10%4%26%66%9%3%22%70%8%2%20%75%6%2%17%38The socioeconomic impacts of employer investments on local communitiesOverall,we decided to base our headline analysis and findings using 1,250 additional job

258、s as the defining threshold for a large investment.This was the level which we felt best balanced the trade-offs outlined previously.As part of our analysis,we have tested the sensitivity of our findings to alternative employment thresholds.In general,this sensitivity analysis implies that our findi

259、ngs are broadly robust to this choice and that the size of the associated socioeconomic impacts gradually increases with the size of the threshold.In the next section we describe characteristics of the investments that were identified in our dataset using this preferred threshold.HISTORICAL LARGE IN

260、VESTMENTSIn this section,we use our dataset to describe the characteristics of“large investments”,as defined in our study,that have taken place in the United States.Overall,the data suggest that between 1990 and 2019,3,161 investments took place,so approximately 109 per year on average.They have bee

261、n unevenly split across US counties.As one might expect,the number of investments that counties receive is very strongly associated with population size(Fig.20)and is highly skewed.For example,in nearly five-in-six counties there were no recorded instances of a large investment.These can be categori

262、zed into two groups as follows:Those that are not included in our investment database.These are counties where there are no instances of a firm adding at least 250 jobs at a single site.In general,these are extremely small(average population in 2021 of 12,600)and typically located in rural areas.Tho

263、se that are included in our investment database.These are counties where there was at least one instance of a firm adding at least 250 jobs at a single site but no instances of a firm adding at least 1,250 jobs.These form part of the“control”group of counties which are used in our model as benchmark

264、 to those counties which did receive investments.They are larger with an average population in 2021 of just over 54,000.On the other hand,the top percentile of counties(those recording 16 or more investments)accounted for nearly half of these investments.They are all large urban areas which attract

265、clusters of large businesses.Fig.20:Average population size of counties depending on number of large investments taking place between 1990 and 2019Average population of county,000sSource:NETS database,Oxford Economics analysis1034-67-1011-15Number of investments in county between 1990 and 201902,000

266、3,0005004,0003,5004,5005,0002,5001,5002501,00016-2021-3031-4041-5039The socioeconomic impacts of employer investments on local communitiesThis distribution of investments is illustrated graphically in Fig.21.Counties where investments have taken place are marked in blue with the shading illustrating

267、 how many have taken place.As shown,there is a relatively strong concentration of investments in the Northeast and Pacific coastal regions of the United States as can be expected given the regional structure of economic activity across the country.Counties that have received investments are plentifu

268、l across the Midwest and the South regions but relatively less prevalent in the West outside of the coastal area.Data sources for socioeconomic variablesInformation on data sources for each indicator and the properties of the associated datasets is contained in Figure 22.As displayed,there was consi

269、derable variation in the time series coverage of some of these indicators.While data on the number of business establishments,unemployment rate,earnings and crime rates stretch back to at least 1990(ideal for our modeling purpose)information on healthcare coverage rates and labor force participation

270、 rates is only available from 2005 and 2008 respectively.On the other hand,these datasets tended to have broadly complete coverage.Fig.21:Location of investments that created at least 1,250 jobs on-site between 1990 and 20190122 to 66 to 1010 to 1515 to 2020 to 3030 to 4040 to 5050 and aboveSource:N

271、ETS database,Oxford Economics40The socioeconomic impacts of employer investments on local communitiesFig.22:Overview of socioeconomic outcome variables tested and data sourcesOutcome indicatorIndicator definitionData sourceTime series coverageCounty coverageBusiness formationNumber of business estab

272、lishmentsCensus Bureau,CBP1990-20203,142Labor market activity%share of working-age population in or actively seeking workCensus Bureau,ACS2005-20213,139Unemployment%share of labor force that are not in workBLS,LAU1990-20213,139Worker prosperityAverage weekly earnings(US$,nominal prices)BLS,QCEW1990-

273、20223,142Healthcare coverage%share of population that have healthcare insurance(overall,private,public)Census Bureau,ACS2008-20213,143Crime rateNumber of crimes recorded per capitaFBI,UCR1990-20203,07541The socioeconomic impacts of employer investments on local communitiesSWNS/Alamy Stock Photo42The

274、 socioeconomic impacts of employer investments on local communitiesAPPENDIX 4:MODELING RESULTS AND SENSITIVITY ANALYSIS Fig.23 and Fig.24 display the time averaged estimated treatment effects(ATTs)with corresponding level of statistical significance.These reflect the average impact of treatment at e

275、ach point both prior and post treatment,with treatment taking place at Event time=0.For example,the contemporaneous impact of large enterprise investment is to reduce the unemployment rate by 0.15 percentage points.This increases to 0.40 percentage points five years after the initial investment.It i

276、s worth noting that each of the coefficients represent the central estimate of the impact at the respective event year.In chapter three,the trajectory of the impact was represented through graphs that displayed the line of best fit i.e.,a smoothed curve.We think that these are likely to represent mo

277、re accurately how the evolution of the impact.For this reason,the five-year impacts shown differ marginally from the respective coefficients in Fig.23 and Fig.24.Fig.23:Time averaged estimated treatment effects(ATTs):unemployment rate,average wages,establishment count,violent crime rateEvent timeUne

278、mployment rate(%)Wages(ln)Establishment count(ln)Violent crime rate(000)(ln)-5-0.0540.0020.006*0.001-40.0310.0070.005*-0.047-3-0.0380.0040.0040.043-2-0.0330.0040.002-0.013-10.0360.0010.002-0.0010-0.154*0.004*0.004*-0.0341-0.259*0.007*0.008*-0.121*2-0.302*0.009*0.014*-0.065*3-0.327*0.013*0.023*-0.087

279、*4-0.356*0.012*0.031*-0.101*5-0.407*0.014*0.036*-0.085*6-0.367*0.016*0.038*-0.0447-0.403*0.016*0.042*-0.094*80.441*0.017*0.050*-0.096*9-0.408*0.017*0.055*-0.099*10-0.2620.016*0.056*-0.089*Statistically significant at 10%level,*Statistically significant at the 5%level,*significant at the 1%level43The

280、 socioeconomic impacts of employer investments on local communitiesFig.24:Time averaged estimated treatment effects(ATTs):private,public and total healthcare coverage rate,labor force participation rate Event timePrivate healthcare coverage(%)Public healthcare coverage(%)Total healthcare coverage(%)

281、Labor force participation(%)-50.016-0.0260.0050.001-4-0.0300.015-0.025-0.013-30.004-0.0010.0000.001-2-0.0070.007-0.0030.000-10.000-0.006-0.002-0.00900.008-0.0030.0010.01310.005-0.0010.0010.0120.010-0.0080.0010.01130.011-0.009-0.0010.02040.018*-0.020*-0.0030.02050.011-0.0070.0030.021*Statistically si

282、gnificant at 10%level,*Statistically significant at the 5%level,*significant at the 1%levelCharacteristic differences across treatment and control groupsCounties in our treatment group tend to be larger than counties in our control group.This is intuitive since treatment assignment is based on an em

283、ployment threshold which reflects a smaller proportion of overall employment for largercounties.Fig.25 illustrates the differential of counties in the control and treatment group across each year of the effective modeling horizon.Fig.25:Distribution of county level population by group at time of tre

284、atmentSource:Oxford EconomicsControlTreatment0100500200700600800900400300Thousands1995ControlTreatment1996ControlTreatment1997ControlTreatment1998ControlTreatment1999ControlTreatment2000ControlTreatment2001ControlTreatment2002ControlTreatment2003ControlTreatment2004ControlTreatment2005ControlTreatme

285、nt2006ControlTreatment2007ControlTreatment2008ControlTreatment2009ControlTreatment2010ControlTreatment2011ControlTreatment2012ControlTreatment2013ControlTreatment2014ControlTreatment2015ControlTreatment201644The socioeconomic impacts of employer investments on local communitiesParallel trends assump

286、tionsA key assumption in the identification of causal ATTs is that the time-path of our outcome variable is conditionally independent of any other factor but exposure to treatment.More formally this is known as parallel trends,which when valid allows us to interpret findings as being causal.Prior to

287、 treatment,macroeconomic trends across treatment and control groups should evolve the same way.If this assumption does not hold,then it would not be possible to interpret any ATT as beingcausal.Visual inspections of the data can be a useful tool to assessing the validity of the parallel trends assum

288、ption across a given dataset.For example,Fig.26 depicts the path of the(unweighted mean)average unemployment rate in the control and the treatment counties that received an investment in 1998 from 1990 to 2018.As illustrated,the two lines are broadly parallel.The model outputs provide a more formal

289、approach to testing for parallel trends.Intuitively,prior to treatment there should be no change in the behavior of our outcome variable across treatment and control groups.This can be more formally tested as pre-treatment coefficients that are statistically insignificant.Unemployment rateSource:Oxf

290、ord Economics19900%2%6%8%10%12%Control group4%Treatment group19911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019Fig.26:Unemployment ratecontrol vs 1998 treatment group45The socioeconomic impacts of employer investments on local communi

291、tiesFig.27 presents results from these tests with the statistical significance of coefficients in pre-treatment years indicated via the standard notation.From these results,it is clear that the impacts on the outcome variables are not statistically significant.The only exception to this across the o

292、utcome variables investigated is the establishment count in years t-5 and t-4.This implies that the assumption of parallel trends is highly valid.Finally,on a related note regarding causality,it is worth highlighting that the Callaway and SantAnna estimator employs an additional propensity weighting

293、 mechanism to match similar counties across groups.This method further strengthens the argument that our ATTs can be interpreted as causal.Fig.27:Pre-treatment test of anticipation effects:unemployment rate,wages,establishment count,violent crime rateEvent timeUnemployment rate(%)Wages(ln)Establishm

294、ent count(ln)Violent crime rate(000)(ln)-5-0.0540.0020.006*0.001-40.0310.0070.005*-0.047-3-0.0380.0040.0040.043-2-0.0330.0040.002-0.013-10.0360.0010.002-0.001*Statistically significant at 10%level,*Statistically significant at the 5%level,*significant at the 1%levelFig.28:Pre-treatment test of antic

295、ipation effects:healthcare coverage rates and labor force participation rateEvent timePrivate healthcare coverage(%)Public healthcare coverage(%)Total healthcare coverage(%)Labor force participation(%)-50.016-0.0260.0050.001-4-0.0300.015-0.025-0.013-30.004-0.0010.0000.001-2-0.0070.007-0.0030.000-10.

296、000-0.006-0.002-0.009*Statistically significant at 10%level,*Statistically significant at the 5%level,*significant at the 1%level46The socioeconomic impacts of employer investments on local communitiesSENSITIVITY ANALYSISModern DiD estimators(such as Callaway&SantAnna)are unable to succinctly captur

297、e the causal impact of multiple treatments on an outcome variable.In other words,these estimates assume that once treated,a unit remains treated indefinitely.A limitation of this approach is that it leads to the samples of investments analyzed in a model such as this to be skewed towards earlier per

298、iods,a fact that is amplified by structural changes in the US economy(the NETS data suggest that the frequency of large direct job creating investments has fallen significantly since the global financial crisis).If the impact of large investments has been largely invariant to time,then these limitat

299、ions will not affect out findings and it would be reasonable to conclude that the results should be broadly representative of the expected local impact of large investment that takes place today.To test this,we have undertaken sensitivity analysis that involved:1.Including counties that are categori

300、zed as multiple treated units.2.Restricting our sample to ensure more recent investments are included produces a similar treatment effect and a portfolio of comparable investments.Including multiple tested unitsOur sample includes 92 counties that are considered multi-treated and therefore exclude f

301、rom modeling.These are counties that tend to be larger than single-treated counties and have seen enterprise investment of at least 1,250+jobs in multiple years.We find that after including these in our models,they produce very similar ATT estimates to the single treated counties.This indicates that

302、 the impact of excluding multi-treated counties is not having an overbearing effect on our ATTs.Fig.29 to Fig.34 compare the estimated ATT coefficient paths for models with and without multi-treated counties.We find the results are relatively insensitive to including multi-treated counties.In genera

303、l,the estimated trajectory of the impact is very similar across all of the outcome metrics investigated as part of our research.Furthermore,of excluding these counties is generally quite small across most models,and all models tend to evolve in the same way;for exampleto reduce the unemployment rate

304、,increase private healthcare coverage,etc.%point change in the unemployed rateSource:Oxford Economics-4-5-3-2-1135-0.45%-0.25%-0.15%-0.10%-0.05%0.00%0.05%0.10%-0.35%Single treated unitsSingle+multi treated units024-0.30%-0.20%-0.40%Fig.29:Unemployment rate ATTestimated impact with and without multi-

305、treated units47The socioeconomic impacts of employer investments on local communitiesFig.33:Labor force participation rate ATTestimated impact with and without multi-treated unitsFig.30:Number of establishments ATTestimated impact with and without multi-treated unitsFig.31:Private healthcare coverag

306、e ATTestimated impact with and without multi-treated unitsFig.32:Average annual earnings ATTestimated impact with and without multi-treated unitsFig.34:Violent crime rate ATTestimated impact with and without multi-treated units%change in the number of establishmentsSource:Oxford Economics-4-5-3-2-11

307、350.0%0.5%2.0%1.5%2.5%3.0%3.5%4.0%Single treated unitsSingle+multi treated units0241.0%point change in private healthcare coverageSource:Oxford Economics-4-5-3-2-1135-4.0%-1.0%-2.0%0.0%1.0%2.0%3.0%Single treated unitsSingle+multi treated units024-3.0%increase in average earningsSource:Oxford Economi

308、cs-4-5-3-2-11350.0%0.2%0.8%0.6%1.0%1.2%1.4%1.6%Single treated unitsSingle+multi treated units0240.4%point change in labour force participationSource:Oxford Economics-4-5-3-2-1135-1.5%0.5%-0.5%1.0%1.5%2.0%2.5%Single treated unitsSingle+multi treated units024-1.0%0.0%change in the violent crime rateSo

309、urce:Oxford Economics-4-5-3-2-1135-14.0%-2.0%-6.0%0.0%2.0%4.0%6.0%Single treated unitsSingle+multi treated units024-8.0%-4.0%-10.0%-12.0%48The socioeconomic impacts of employer investments on local communitiesRestricting our sample to include more recent investmentsAs described,assigning counties be

310、tween control and treatment groups using the full sample of data means our sample is weighted more towards investments during the 1990s.This is beneficial from the perspective of typically having a longer post-treatment period available to evaluate the impact but runs the risk that our results are n

311、ot representative of more modern investments.45 It is worth noting that this restriction would fundamentally alter the composition of both the control and treatment groups.Some counties that were previously in our treatment group will have moved to the control group,and some counties discarded as be

312、ing multi-treated will now appear as single treated.To ensure comparability,we have maintained a singular control group in both cases and restricted the treatment group to post-2000 events.In theory,this might be a particular concern given changes to the structure of the US economy during the interv

313、ening period.This is manifested in the change in the industrial composition of investments over time.For example,Fig.35 compares the share of single treatment investments by sector in our full sample and a sample that is restricted to post-2000.Consistent with priors,in the latter,there is a signifi

314、cant fall in the share of investments in the manufacturing sector(34%to 24%)with a compensating increase in parts of the services sector notably healthcare,transport and communications,and warehousing and storage.To formally test this,we restrict our sample to only include investments from 2000 onwa

315、rds and re-run the model.This year was chosen as it removed approximately 50%of the single-treatment events that are used as the basis for our core analysis.The results from this exercise are displayed in Fig.36 to Fig.39.N.B.,since the models used to assess the impact of large investments on the la

316、bor force participation rate and healthcare insurance rates use an exclusively post-2000 sample of investments,this restriction has no impact on the estimated results and hence we do not report them here.45 Fig.35:Comparing the share of broad sector investments across our full and restricted sampleS

317、ource:Oxford EconomicsAgriculture and mining0%5%B2C manufacturingIndustrial manufacturingTransport equipmentTransport,communicationsand utilitiesWarehousing,storage and retailFinancial andbusiness servicesHospitality and leisureHealthcareMiscellaneousConstruction10%15%20%25%4%2%5%14%16%13%4%6%5%8%9%

318、10%18%18%2%2%14%22%13%13%Full sampleRestricted sample49The socioeconomic impacts of employer investments on local communitiesEach figure below contains our full sample post-treatment model estimates with confidence bounds,overlayed with the corresponding estimates from the restricted sample.Fig.38:A

319、nnual earnings ATT using a restricted sampleFig.36:Unemployment rate ATT using a restricted sample(pp change)Fig.37:Number of establishments ATT using a restricted sampleFig.39:Violent crime rate ATT using a restricted sample%point change in unemployment rateSource:Oxford Economics estimates102347-0

320、.8%-0.4%-0.6%-0.3%-0.2%-0.1%0.0%Full sampleRestricted sample5-0.7%-0.5%Years from investment,0=year of investment6%change in number of establishmentsSource:Oxford Economics estimates1023470%3%1%4%5%6%7%Full sampleRestricted sample52%Years from investment,0=year of investment6%change in annual earnin

321、gsSource:Oxford Economics estimates102347-0.5%1.0%0.0%1.5%2.0%2.5%3.0%Full sampleRestricted sample50.5%Years from investment,0=year of investment6%point change in the violent crime rateSource:Oxford Economics102347-25%-15%-10%-5%0%5%Full sampleRestricted sample5-20%Years from investment,0=year of in

322、vestment650The socioeconomic impacts of employer investments on local communitiesIn all cases,restricted model estimates tend to fall within the same distributive bounds of our original model.This suggests that both models come from the same data generating process.We can test this more formally usi

323、ng a t-test across both ATT estimates.Fig.40:2-tailed t-test of ATTs across full and restricted sample modelsModelP-valueOutcomeUnemployment model0.70Cannot reject the null hypothesisNumber of establishments0.72Cannot reject the null hypothesisAverage earnings0.60Cannot reject the null hypothesisCri

324、me rate0.40Cannot reject the null hypothesisNote:Healthcare and Labor Force Participation models use a smaller sample and results are unaffected by restricting the sample.51The socioeconomic impacts of employer investments on local communitiesOXFORD ECONOMICSOxford Economics was founded in 1981 as a

325、 commercial venture with Oxford Universitys business college to provide economic forecasting and modelling toUK companies and financial institutions expanding abroad.Since then,we have become one of the worlds foremost independent global advisory firms,providing reports,forecasts and analytical tool

326、s on more than 200 countries,100industries,and 8,000 cities and regions.Our best-in-class global economic and industry models and analytical tools give us an unparalleled ability to forecast external market trends andassess their economic,social and business impact.Headquartered in Oxford,England,wi

327、th regional centers in New York,London,Frankfurt,and Singapore,OxfordEconomics has offices across the globe in Belfast,Boston,Cape Town,Chicago,Dubai,Dublin,Hong Kong,Los Angeles,Mexico City,Milan,Paris,Philadelphia,Stockholm,Sydney,Tokyo,and Toronto.We employ 450 staff,including more than 300profes

328、sional economists,industry experts,and business editorsone of the largest teams of macroeconomists and thought leadership specialists.Our global team is highly skilled in a full range of research techniques and thought leadership capabilities from econometric modeling,scenario framing,and economic i

329、mpact analysis to market surveys,case studies,expert panels,and web analytics.Oxford Economics is a key adviser to corporate,financial and government decision-makers and thought leaders.Our worldwide client base now comprises over 2,000 international organizations,including leading multinational com

330、panies and financial institutions;key government bodies and trade associations;and top universities,consultancies,and think tanks.Released for internal use in December 2023Released for internal use in December 2023All data shown in tables and charts are Oxford Economics own data,except where otherwi

331、se stated and cited in footnotes,and are copyright Oxford Economics Ltd.Amazon,Inc.provided the funding for this independent study.The findings of this research reflect the analysis of Oxford Economics.This report is confidential to Amazon and may not be published or distributed without their prior

332、written permission.The modeling and results presented here are based on information provided by third parties,upon which Oxford Economics has relied in producing its report and forecasts in good faith.Any subsequent revision or update of those data will affect the assessments and projections shown.T

333、o discuss the report further please contact:Laurence Wilse-Samson Oxford Economics 5 Hanover Sq,8th Floor New York,NY 10004Tel:+1 646-786-1879Cover PhotoPetinov Sergey Mihilovich/SGlobal headquartersOxford Economics Ltd Abbey House 121 St Aldates Oxford,OX1 1HBUKTel:+44(0)1865 268900London4 MillbankLondon,SW1P 3JA UKTel:+44(0)203 910 8000FrankfurtMarienstr.15 60329 Frankfurt am Main GermanyTel:+49