盡管就業(yè)和產(chǎn)出的相對比重長期下降，但美國制造業(yè)已從2008-9年的衰退中大幅反彈。事實上，這是自20世紀80年代初以來該行業(yè)首次完全恢復其產(chǎn)出份額的衰退（Celasun等人。2014年）。在國際上，美國約占全球制造業(yè)的20%，與中國大致相同。盡管國內(nèi)制造業(yè)的就業(yè)還沒有完全恢復到衰退前的水平，但該行業(yè)仍然占到該國高工資藍領(lǐng)工作崗位的一半以上（麥肯錫全球研究所，2012年）。制造業(yè)復蘇的同時，天然氣價格也大幅下跌（EIA2013A）。不斷下降的生產(chǎn)成本，特別是來自頁巖地層的生產(chǎn)成本，極大地增加了產(chǎn)量和估計儲量，降低了消費者的天然氣價格。2007年至2012年間，實際天然氣價格下降了約50%。2005年以來，國內(nèi)天然氣產(chǎn)量增長50%，2016年達到2.8萬億立方英尺（tcf）。僅在2011年的一年時間里，美國能源信息署（EIA）就將其技術(shù)上可開采頁巖氣的估計值翻了一番多，達到827tcf。隨后，對估算值進行了多次上調(diào)，最近的估算表明，美國可能擁有超過1100 tcf的技術(shù)可采頁巖氣資源（EIA 2013b）。結(jié)合其對非頁巖氣資源的估計，EIA預計美國的技術(shù)可采天然氣儲量總體超過100年（EIA 2017）。很少有專家質(zhì)疑美國經(jīng)濟正從天然氣產(chǎn)量增加和價格下降中受益。國內(nèi)生產(chǎn)總值增長、天然氣生產(chǎn)部門擴大就業(yè)、降低電力和天然氣價格、改善貿(mào)易平衡以及增加稅收，這些都是共同體現(xiàn)的收益。制造業(yè)直接將天然氣作為燃料，并通過發(fā)電間接消耗天然氣。電力和制造業(yè)都大幅增加了天然氣消費，住宅和商業(yè)部門幾乎沒有增長。根據(jù)標準的微觀經(jīng)濟學理論，天然氣等投入品價格的下降會降低邊際生產(chǎn)成本，增加產(chǎn)出，降低產(chǎn)出價格。產(chǎn)量的增加增加了就業(yè)，而天然氣價格相對于工資的下降可以增加或減少就業(yè)，這取決于天然氣和就業(yè)在生產(chǎn)中是互補的還是替代的。就業(yè)增長的規(guī)模、地點和時機尚不清楚，是政策界爭論不休的話題。這一點尤其適用于對個別行業(yè)或國家特定地區(qū)的微觀考察。天然氣價格走低也引發(fā)了有關(guān)未來天然氣出口的政策辯論。與其他國家相比，美國的低價格增加了天然氣出口的盈利能力，無論是通過管道還是通過液化形式的船舶。與此同時，天然氣出口增加將提振美國價格，有利于天然氣生產(chǎn)商，但損害消費者，特別是天然氣密集型行業(yè)。這些相反的影響是就應允許出口快速增長還是應限制增長展開激烈政策辯論的基礎(chǔ)。最近的幾項研究試圖估計最近天然氣價格下跌對制造業(yè)就業(yè)的影響。這些估計表明，天然氣價格下跌帶來的制造業(yè)就業(yè)增長可能相當大，如www.rff.org | 1 Resources for the Future | Gray，Linn，在能源密集度最高的十分之一的行業(yè)中，摩根斯坦恩高達9.1%，在能源密集度最高的行業(yè)（化肥制造業(yè)）中，摩根斯坦恩高達30%。1，然而，這些論文依賴于聚集的數(shù)據(jù)，并幾乎沒有注意到其他因素可能與天然氣價格，也影響就業(yè)，如熟練的勞動力可用性，接近中間輸入，或其他位置特定的變量。為了評估天然氣價格對制造業(yè)復蘇的影響，并描述擴大天然氣出口的潛在影響，我們試圖在本文的最新文獻基礎(chǔ)上取得一些進展。首先，我們分析了天然氣價格對地方（縣級）就業(yè)的影響，定義在高度細分的工業(yè)層面。這一細節(jié)有助于分析天然氣價格下跌的總體和地理影響，而其他研究則依賴于國家一級的數(shù)據(jù)?？h級數(shù)據(jù)的使用允許我們進行第二個前進，即考慮其他位置特定的因素，例如接近關(guān)鍵輸入。這種方法建立在集聚和產(chǎn)業(yè)動態(tài)的文獻基礎(chǔ)上（例如，Ellison和Glaeser 1997）。使用匯總數(shù)據(jù)無法解釋這些因素，這些因素最終與天然氣價格相關(guān)。因此，對特定地點因素的控制大大減少了觀察到的天然氣價格下降對就業(yè)的估計影響。我們使用機密的工廠級普查信息來開發(fā)一個簡化的面板數(shù)據(jù)模型，將制造業(yè)就業(yè)與天然氣和電價聯(lián)系起來，控制1這些計算不包括與提供直接用于石油和天然氣生產(chǎn)的材料的行業(yè)相關(guān)的間接影響，比如增加了鋼廠的產(chǎn)量，以生產(chǎn)鉆機上使用的套管。對于其他可能影響增長的因素，并估計長期（五年）影響。我們發(fā)現(xiàn)，天然氣價格對縣級就業(yè)的影響在統(tǒng)計學上是顯著的，其影響通常隨著天然氣強度的增加而增加。與卡恩和曼蘇爾（2013）等先前研究的結(jié)果一致，電價也影響就業(yè)，這表明天然氣價格直接和間接影響制造業(yè)就業(yè)。與其他研究類似，我們的估計包括天然氣價格變化引起的天然氣和勞動力之間的生產(chǎn)擴張和替代的影響。相應地，所得到的模型估計值被用來模擬2007年至2012年間50%的天然氣價格下降所帶來的就業(yè)增長。我們發(fā)現(xiàn)，天然氣價格的下跌使制造業(yè)的整體就業(yè)率上升了0.6%。這些對天然氣價格引起的收益的估計比最近其他研究報告的要小，這至少部分是由我們對聚集因素的控制所解釋的。相對于觀察到的制造業(yè)就業(yè)總量的波動，我們估計的就業(yè)影響也很?。?007年至2012年間，制造業(yè)就業(yè)總量下降了17.8%，2010年至2012年后衰退期間，制造業(yè)就業(yè)總量增長了6.1%。盡管天然氣價格對制造業(yè)就業(yè)總量的影響不大，但對天然氣密集型行業(yè)的就業(yè)確實產(chǎn)生了實質(zhì)性影響。2007年至2012年期間，天然氣價格的下降使相對天然氣密集型行業(yè)（即，位于www.rff.org | 2未來資源| Gray、Linn和Morgenstern天然氣密度分布前四分之一的行業(yè)）的就業(yè)率增加了1.8%，這比整個行業(yè)的估計值高出三倍制造業(yè)，反映了天然氣作為天然氣密集型產(chǎn)業(yè)投入的重要性。與總體估計一樣，這一估計值小于其他近期研究報告中的估計值。2在2007-12年和2010-12年期間，天然氣密集型行業(yè)的就業(yè)增長率始終高于制造業(yè)平均水平?？偟膩碚f，天然氣價格的下降解釋了天然氣密集型行業(yè)就業(yè)增長速度加快一半以上的原因。因此，天然氣價格只解釋了這一時期就業(yè)總量變化的一小部分，但在就業(yè)增長中跨行業(yè)變化的很大一部分。預計天然氣出口的增長對天然氣價格的影響相對較小，將在3%至9%之間提高（EIA 2014）。因此，對我們模型的模擬表明，出口增加對整個制造業(yè)就業(yè)的影響相對較小，所有行業(yè)的就業(yè)減少0.1%至0.3%，這取決于出口引起的價格變化的幅度。對于天然氣密集型行業(yè)，就業(yè)率將下降0.2%至0.5%。本文第2節(jié)回顧了國內(nèi)天然氣價格與制造業(yè)就業(yè)之間聯(lián)系的最新文獻。第3節(jié)和第4節(jié)描述了建模設置和工廠級數(shù)據(jù)。模型估計和仿真結(jié)果見第5節(jié)和

Despite long-term declines in its relative share of employment and output, US manufacturing has rebounded sharply from the 2008–9 recession. In fact, this is the first recession where the sector has fully recovered its share of output since the early 1980s (Celasun et al. 2014). Internationally, the United States represents about 20 percent of global manufacturing—roughly the same share as China. Even though employment in domestic manufacturing has not fully returned to prerecession levels, the sector still accounts for more than half of the high-wage blue-collar jobs in the country (McKinsey Global Institute 2012).  The recovery of the manufacturing sector has coincided with a sharp drop in natural gas prices (EIA 2013a). Declining production costs, particularly from shale formations, have dramatically increased output and estimated reserves and reduced gas prices for consumers. Between 2007 and 2012, real natural gas prices declined by about 50 percent. Since 2005, domestic natural gas production has increased by 50 percent, reaching 2.8 trillion cubic feet (tcf) in 2016. In one year alone, 2011, the US Energy Information Administration (EIA) more than doubled its estimates of technically recoverable shale gas to 827 tcf. Subsequently, estimates have been revised upward several times, with recent estimates suggesting the United States may have more than 1,100 tcf of technically recoverable shale gas resources (EIA 2013b). Combined with its estimates of non–shale gas resources, EIA projects that the United States has more than 100 years of technically recoverable natural gas reserves overall (EIA 2017).  Few experts dispute that the US economy is benefiting from higher natural gas production and lower prices. GDP growth, employment expansion in the natural gas production sector, lower electricity and natural    gas prices, improvements in trade balance, and increases in tax revenues are all among the commonly articulated gains. The manufacturing sector consumes natural gas directly as a fuel and indirectly through its use in electricity generation. Both the electricity and manufacturing sectors have increased their gas consumption substantively, with few gains in the residential and commercial sectors.  Based on standard microeconomic theory, a decline in the price of an input, such as natural gas, reduces marginal production costs, increases output, and reduces output prices. The output increase raises employment, and the decrease in natural gas price relative to wages can increase or decrease employment depending on whether natural gas and employment are complements or substitutes in production. The size, location, and timing of the job gains are unclear and are the subject of considerable debate in policy circles. This is particularly true when looking granularly at individual industries or at particular regions of the country.  Lower natural gas prices have also generated a policy debate about future natural gas exports. Low prices in the United States relative to other countries have increased the profitability of exporting natural gas, either by pipeline or by ship in liquefied form. At the same time, an increase in gas exports would boost US prices, benefiting natural gas producers but harming consumers, particularly gas-intensive industries. These opposing effects are the basis for a vigorous policy debate over whether exports should be allowed to grow rapidly or growth should be restrained.  Several recent studies have attempted to estimate the effects of the recent decline in natural gas prices on manufacturing employment. These estimates suggest that the manufacturing employment gains from the gas price declines are potentially quite large—as     www.rff.org  |  1  Resources for the Future  |  Gray, Linn, and Morgenstern    much as 9.1 percent for industries in the top decile of energy intensity and up to 30 percent for the most energy-intensive industry (fertilizer manufacturing). 1 However, these papers rely on aggregated data and pay virtually no attention to other factors that may be correlated with natural gas prices and also affect employment, such as skilled labor availability, proximity to intermediate inputs, or other location-specific variables.  To assess the effects of natural gas prices on the recovery of the manufacturing sector and to characterize the potential effects of expanding natural gas exports, we attempt to make several advances over the recent literature in this paper. First, we analyze the effects of natural gas prices on local (county-level) employment, defined at a highly disaggregated industrial level. This detail enables an analysis of the aggregate and geographic effects of natural gas price declines, in contrast to the other studies, which have relied on national-level data. The use of county-level data allows us to make a second advance—namely, the consideration of other location-specific factors, such as proximity to key inputs. This approach builds on the literature on agglomeration and industry dynamics (e.g., Ellison and Glaeser 1997). Accounting for these factors is not possible using aggregated data, and these factors turn out to be correlated with natural gas prices. Consequently, controlling for the location-specific factors substantially reduces the estimated effects of the observed natural gas price declines on employment.  We use confidential plant-level Census information to develop a reduced -form panel data model linking manufacturing employment to natural gas and electricity prices, controlling    1These calculations exclude indirect effects associated with industries providing materials directly used in oil and gas production, such as the increased output of steel mills to produce the casing used on drilling rigs.    for other factors that may affect growth, and estimate long-term (five-year) effects. We find that natural gas prices have a statistically significant effect on county-level employment, with effects typically increasing with natural gas intensity. Consistent with the findings of previous studies such as Kahn and Mansur (2013), electricity prices also affect employment, which suggests that natural gas prices affect manufacturing employment both directly and indirectly. Similarly to other studies, our estimates include the effects of both expanding production and substitution between natural gas and labor induced by changes in natural gas prices.  The resulting model estimates, in turn, are used to simulate the employment gains arising from the 50 percent natural gas price decline that occurred between 2007 and 2012. We find that the drop in natural gas prices raised overall manufacturing employment by 0.6 percent. These estimates of the gas price–induced gains are smaller than those reported in other recent studies, which is at least partly explained by our controlling for the agglomeration factors. Our estimated employment impacts are also small relative to the observed swings in aggregate manufacturing employment over the period: either the aggregate decline of 17.8 percent between 2007 and 2012 or the aggregate increase of 6.1 percent in the post-recession period between 2010 and 2012.  Although natural gas prices had modest effects on aggregate manufacturing employment, they did have substantial effects on employment in gas -intensive industries. The decline in natural gas prices between 2007 and 2012 increased employment by 1.8 percent in the relatively gas -intensive industries (i.e., those in the top quartile of the     www.rff.org  |  2  Resources for the Future  |  Gray, Linn, and Morgenstern    gas intensity distribution), which is three times greater than the estimate for the entire manufacturing sector, reflecting the importance of natural gas as an input for the gas-intensive industries. Like the aggregate estimate, this estimate is smaller than that reported in other recent studies.2 In both the 2007–12 and 2010–12 periods, the gas-intensive industries experienced consistently higher employment growth than the manufacturing average. Overall, the decline in natural gas prices explains more than half the faster employment growth rates observed in the gas-intensive industries. Thus natural gas prices explain only a small portion of the aggregate employment changes but large shares of cross-industry variation in employment growth during these time periods.  Increases in natural gas exports are expected to have relatively small impacts on natural gas prices, raising them somewhere between 3 and 9 percent (EIA 2014). Consequently, simulations of our model suggest that higher exports would have relatively small impacts on overall manufacturing employment, reducing employment by 0.1 to 0.3 percent across all industries, depending on the magnitude of the export-induced price change. For the gas-intensive industries, the employment reductions would be 0.2 to 0.5 percent.  Section 2 of the paper reviews the recent literature on the link between domestic natural gas prices and manufacturing employment. Sections 3 and 4 describe the modeling setup and plant-level data. The model estimates and simulation results are presented in sections 5 and