We evaluate the employment effect of the green part of the largest fiscal stimulus in recent history, the American Recovery and Reinvestment Act (ARRA).
65 pages
100 KB – 65 Pages
PAGE – 1 ============
CLEAN ECONOMY WORKING PAPER SERIES THE EMPLOYMENT IMPACT OF GREEN FISCAL PUSH: EVIDENCE FROM THE AMERICAN RECOVERY ACT David Popp Syracuse University National Bureau of Economic Research Francesco Vona OFCE Sciences-Po and SKEMA Business School, France CMCC Ca™Foscari, Italy Giovanni Marin University of Urbino Carlo Bo, Italy Ziqiao ChenSyracuse University
PAGE – 2 ============
1 The Employment Impact of Green Fiscal Push: Evidence from the American Recovery Act David Popp * Francesco Vona ƒ Giovanni Marin ⁄ Ziqiao Chen § May 2 8, 2020 Abstract We evaluate the employment effect of the green part of the largest fiscal stimulus in recent history, the American Recovery and Reinvestment Act (ARRA). Each $1 million of green ARRA created 15 new jobs that emerged especially in the post -ARRA period (2013 -2017). We find little evidence of significant short -run employment gains. Green ARRA creates more jobs in commuting zones with a greater prevalence of pre -existing green skills. Nearly half of the jobs created by green ARRA investments were in construction or waste management. Nearly all new jobs created are manual labor positions. Noneth eless, manual labor wages did not increase . Keywords: employment effect, green subsides, American Recovery Act, heterogeneous effect, distributional impacts JEL Codes: E24, E62, H54, H72, Q58 Acknowledgements: This project has been supported in part th rough the Smart Prosperity Institute Research Network and its Greening Growth Partnership, which is supported by a Social Sciences and Humanities Research Council of Canada Partnership Grant (no. 895 -2017-1018), as well as by Environment and Climate Change Canada™s Economics and Environmental Policy Research Network (EEPRN). This work was also supported by Horizon 2020 Framework Programme, project INNOPATHS [grant number 730403]. We thank seminar participants at the University of Bremen, APPAM, and Greening Growth Partnership & Economics and Environmental Policy Research Network Annual Symposium for helpful comments. * Syracuse University, United States; NBER. ƒ OFCE Sciences -Po; SKEMA Business School, France and CMCC Ca™ Foscari, Italy. ⁄ University of Urbino Carlo Bo, Italy. § Syracuse University, United States.
PAGE – 3 ============
2 I. Introduction The effect of environmental policy on employment is still hotly debated and polarized , with advocates on both sides ignor ing or exaggerat ing the labor market costs and benefits of environmental regulations . Advocates of stronger environmental policies argue that such policies create high -paying figreen jobsfl, while critics point to the job losses in energy -intensive industries that they are sure will follow. Previous literature finds that net effect of environmental policies on employment is small especially when general equilibrium effects and offsetting mechanisms are accounted for ( Morgenstern et al., 2002; Hafstead and Williams, 2018 ; Metcalf and Stock, 2020 ). However, other studies find job losses concentrated in polluting industries (Greensto ne, 2002, Kahn and Mansur, 2013) and among unskilled workers (Yip, 2018; Marin and Vona, 2019). Adverse impacts on manual labor are o f particular concern for policy -makers , given the secular decline in their employability and wages driven by automation and globalization (Autor et al., 2003; Autor et al. , 2013). While the previous literature has evaluated the effect of policies imposing a cost on pollution (either through standards or prices) on labor markets , less attention has been devoted to the potentia l of green subsidies opening up new employment opportunities in the so-called green economy. Our paper informs the burgeoning policy debate on green fiscal plans , by focusing on the evaluation of a big push for the green economy, namely the green part of t he American Recovery and Reinvestment Act (ARRA, henceforth). The full stimulus package included over $350 billion of direct government spending, and an additional $260 billion in tax reductions ( Aldy, 2013). We focus on the direct spending targeted at green investments, which constituted approximately 17% of all direct government spending in ARRA. Examples of such spending include Department of Energy (DOE) block grants to states to support energy efficiency aud its
PAGE – 4 ============
3 and retrofits, investments in public transport and clean vehicles, and Environmental Protection Agency (EPA) spending to clean up brownfield sites. Because a large share of green spending was devoted to public investments, green ARRA may have a cumulat ive effect stretching beyond the stimulus period (Council of Economic Advisers, 2013 , 2014). We thus differentiate between the short – and long -term effect of green ARRA. We evaluate the employment gains triggered by the green stimulus, its heterogeneous ef fect depending on the level of local green capabilities and the way in which the green stimulus has affected different sectors and groups of workers. Our evaluation is timely and important as p roposals for green stimuli investments have attract ed a great deal of attention , both as part of possible recovery packages after COVID -19 lockdowns and as part of Green New Deal plans proposed by the European Commission , the International Energy Agency, the International Monetary Fund and some D emocrats in the US (He lm, 2020) . Our analysis makes three contributions to the discussion of heterogeneous labor market effects. First, using data on green skills from Vona et al. (2018), we show that the effectiveness of green investments varies depending on the pre -existing skill base of a community. Second, we estimate the effects of green ARRA investments on different sectors and sets of occupations to identify those workers receiving the most benefits from green investments . Third, our focus on heterogeneous effects across different types of workers also adds to the literature on structural transformations and inequality in local labor markets (e.g ., Autor et al. , 2013; Acemoglu and Restrepo, 2020). A key difference between in vestments in the green economy , especially in building retrofitting and energy infrastructures, and in automation is that the former increase the relative demand of manual workers, while the latter decreases it. This implies that manual workers that are displaced by carbon pricing policies in energy i ntensive sectors ( Marin and Vona, 2019; Yip, 2019 ) may find new employment opportunities in sectors related to the green economy, such
PAGE – 5 ============
4 as con struction and waste management. Our research considers whether green investments can facilitate this transition in local labor markets. Our analysis also contributes to the broader literature estimating the effects of the 2009 Recovery Act. We add to the empirical literature on fiscal multipliers looking at the effect of a type of spending, i.e. in the green economy, that will become increasingly important in the future (see Chodorow -Reich, 2019 for a survey). In the spirit of recent contributions seeking to isolate the microeconomics mechanisms of the local multiplier (e.g. Moretti, 2010; Garin, 2018; Dupor and McCror y, 2018; Auerbach et al., 2019 ), we study the time profile of the effect, the role of key mediating factors and some mechanisms through which the green stimulus impact on the local economy. Prev ious literature on other aspects of the Recovery Act exploit geographical variation in expenditures and isolate its exogenous component , and thus a causal effect, using pre -existing formulas to allocate federal funds (Wilson, 2012; Chodorow -Reich et al., 2012; Nakamura and Steinsson, 2014; Dupor and Mehkari, 2016; Chodorow -Reich, 2019 ). However, identifying the causal effect of the green stimulus presents three additional challenges. First, the green stimulus is small relative to the non -green stimulus. Controlling for non -green ARRA expenditures is essential, but po tentially introduces another endogenous variable complicating the identification of the green ARRA effect ( Angrist and Pischke, 2008 ). The trade -off is between an error of misspecification from not including non -green ARRA and a bias in estimating the gree n ARRA effect for including a bad control (non -green ARRA) correlated with the error term. We address the first challenge by including a set of twenty dummies representing each vigintile of per capita non-green ARRA. This allows us to compare the effect of green ARRA in communities that received similar levels of non -green ARRA investments and to test the robustness of our results to
PAGE – 6 ============
5 the exclusion of vigintiles in which the dispersion of green ARRA spending is very hig h or low or for which the correlation between green and non -green ARRA is very high. Second, the allocation of green investments may be dependent on characteristics of the local economy. In general, ARRA spending targeted areas hardest hit by the recessio n and is endogenous by construction. The share of ARRA that is green may be further influenced by features of the economy specific to green investments, such as the presence of a federal DOE laboratory or the renewable energy potential of a region. We addr ess these concerns through two sets of control variables. The first set captures the economic conditions in commuting zone before the great recessions and are quite standard in the literature evaluating the Recovery Act (e.g. Wilson, 2012; Chodorow -Reic h et al., 2012; Chodorow -Reich, 2019). The second set of controls are specific to the green economy, such as the stringency of environmental regulation in the local area (Greenstone, 2002), wind and solar energy potential (Aldy, 2013) and the pre -existing base of green skills in each commuting zone (Vona et al., 2018). Third, we observe that even after controlling for these observables, areas receiving more green ARRA experienced higher employment growth before the great recession. While standard state or r egional fixed effects are sufficient to eliminate the pre -trend for non -green ARRA investments, they do not eliminate the pre -trend on total employment for green ARRA. We address these pre -trends in two ways. First, we allow the effect of green ARRA invest ments to vary across three periods: the pre -ARRA period (2005 -2007); the short -term (2009 -2012) and the long -term (2013 -2017). We compute the long – and short -run effect of green ARRA by subtracting its effect before 2008. Second, we use a standard shift -share instrument (e.g., Nakamura and Steinsson, 2014), where we combine the pre -sample share of different types of green spending in each commuting zone with the green ARRA shift. While neither solution is perfect, comparing the OLS
PAGE – 8 ============
7 effects. While the unavoidable presence of pre -trends prevents us from drawing firm conclusions on the overall effect of green AR RA, its impact becomes much clearer when we explore several dimensions of heterogeneity , for which pre -trends are less of a concern . Green ARRA creates more jobs in commuting zones with a greater prevalence of pre -existing green skills. As the presence of green skills in a community is also strongly correlated with the allocation of green ARRA subsidies, our results provide evidence of the green stimulus as a successful example of picking winners. Looking at specific sectors of the economy, we see the poten tial of a green stimulus to reshape an economy and increase the local demand for green tasks . Nearly half of the jobs created by green ARRA investments were in construction or waste management. Nearly all of the new jobs created are manual labor positions. Importantly, w hile we find evidence of pre -trends when evaluating total employment, we find no evidence of pre -trends when we study heterogeneous impact s across sectors and workers , providing us with confidence that our results a re credible . Even though the largest employment gains were for manual laborers with at least some college education, manual labor wages did not increase. These missing wage gains may either reflect the fact that the green stimulus was too small to o ffset the long -term deterioration of the bargaining power of manual workers, or the poor quality of the jobs created. While further research is required to understand the impact of green subsidies on labour market inequalities, these results suggest that t he green stimulus may create new opportunities for those most affected by globalization and automation. The remainder of the paper is organized as follows. Section 2 gives the necessary background on the green part of the Recovery Act. Section 3 presents the data used for this project as well as preliminary descriptive statistics. Section 4 discusses the empirical strategy, while Section 5 the main results. Section 6 discusses the policy implications of our study .
PAGE – 9 ============
8 II. The Green component of the Recovery Act In response to the Great Recession, the American Recovery and Reinvestment Act (ARRA) of 2009, commonly known as the stimulus package, invested over $800 billion in the forms of tax incentives and federal spending programs to stimulate the US economy. Throu gh ARRA spending programs, federal agencies partnered with state and local governments, non -profit and private entities to help fiput Americans back to workfl. Naturally, much of the spending programs funded projects that provide immediate job opportunities, such as highway construction, or filled state budget shortfalls to bail out the school system and save the jobs of teachers and school staff. Figure 1 shows the breakdown of funds by federal agency, which confirms large ARRA spending on education and tran sportation. Figure 1 ΠARRA spending by awarding Department / Agency Notes : own elaboration based on Recovery.gov data from NBER data repository.
PAGE – 10 ============
9 While the primary goal of ARRA was to stimulate macroeconomic growth and provide job opportunities, part of the funds were invested in fi– environmental protection, and infrastructure that will provide long -term economic benefitsfl (American Recovery and Rei nvestment Act of 2009). These include both direct spending intended for immediate job creation, su ch as Department of Energy spending for renewable energy and energy efficiency retrofits and Envir onmental Protection Agency grants for brownfield redevelopme nt, as well as tax breaks and loan guarantees for renewable energy. Our work focuses on the impact of direct spending intended for job creation, asking both whether these green investments stimulated employment and what types of workers may benefit from a green stimulus. Among the key principles motivating infrastructure investments in ARRA was that facilitating the transition to energy efficient and clean energy economy would lay the foundation for long -term economic growth ( Office of the Vice President , 2010). As a result, ARRA included more than $90 billion for clean energy activities, including $32.7 billion in Department of Energy contracts and grants to support projects such as energy efficiency retrofits, the development of renewable energy resources, public transport and clean vehicles, and modernizing the electric grid (Aldy, 2013). To meet the Obama administration™s target of doubling renewable energy generation by 2012, DOE provided assistance for a large number of projects related to renewable ene rgy; for example, the Massachusetts Clean Energy Center received $24.8 million to design, construct and operate a wind turbine blade testing facility ( Department of Energy , 2010). Moreover, $3.4 billion in cost -shared grants supported the deployment of smart grid technology, generating more than $4.5 billion of co -investment (Aldy 2013). ARRA funding also supported the expansion of the Weatherization Assistance Program, which supports low -income families for energy efficiency improvements (Fowlie et al., 2018).
PAGE – 11 ============
10 The Environmental Protection Agency (EPA) oversaw most ARRA programs designated for environmental protection. The largest of these programs was $6.4 billion for Clean and Drinking Water State Revolving Funds, which are among the programs analyzed i n Dupor and McCrory (2018). An additional $600 million was set aside for EPA™s Superfund pro gram to clean up contaminated sites such as the New Bedford Harbor site in Massachusetts and the Omaha Lead Site in Nebraska, to which the EPA allocated $30 million and $25 million, respectively 6 (Office of the Vice President, 2010 ). Another $200 million was invested in the Leaking Underground Storage Tank Trust Fund for the prevention and cleanups of leakage from underground storage tanks. Other EPA funds were alloc ated to improvements of infrastructures such as wastewater treatment facilities and diesel emissions reduction (Environmental Protection Agency, 2009). A. Data on ARRA awards Our analysis covers the universe of contracts, grants and loans awarded under the AR RA between 2009 and 2012. Recipients of ARRA funding are required to submit reports through FederalReporting.gov , which include information on the amount of expenses and the description of projects. 7 We retrieved data from FedSpending.org on these records derived from reports submitted by non -federal entities who received ARRA funding. In line with most recent evaluations of ARRA ( Dupor and Mehkari, 2016; Dupor and McCrory, 2018), our unit of analysis is the local labor market, i.e. the so -called commuting zone (CZ). We aggregate county -level data into 709 Commuting Zones based on the official CZ definitions from the 2000 Decennial Census. As in Dupor and Mehkari (2016), we exclude 122 6 Information on active and archived Superfund sites is available at https://cumulis.epa.gov/supercpad/ cursites/srchsites.cfm , last accessed May 27, 2020. 7 This website is no longer use, but archived data are available at https://data.nber.org/data/ARRA/ , last accessed March 6, 2020.
100 KB – 65 Pages