ANNUAL REVIEWS noo. Rev, Resour Econ 2018, 10381-404 “The Arma! Reie f Rem Eames online Ips. on/10.1146/acmarer-resouree- 109si7-casi94 Copyighe © 2018 by Anal Revews Al righ rsered Corresponding author JEL coder 013, 018, Qt, Q53, 056 pe cnnE ae PRADEEP XEROX CENTER HINDU COLLEGE & K.M.COLLEGE MOBILE: 8130462424 i ee ee | Annual Review of Resource Economics Efficacy of Command-and- Control and Market-Based Environmental Regulation in Developing Countries Allen Blackman,'* Zhengyan Li,? and Antung A. Liu? Keywords ‘command-and-control, market-based instruments, developing country Abstract Like thei counterpars in indstialzed countries, environmental regsla~ tors in developing counties rely principally on two types of instruments: «ommand-and-control (CAC) policies, such as emissions and technology standards, and to a lesser extent, market-based instruments (MBI), such as emissions fees and radable permits. Butthese regulators oftenackthe capac- ity o implement, monitor, and enforce CAC and MBI polices. As a result, the elcacy of those policies is an empirical mater, We review emerging experimental and quasi-experimental evidence on CAC and MBI policies in developing countries, specifically, from 32 swdies of CAC policies and 8 studies of MBIs. Although drawn from a small and decidedly nonrandom sample of countries and policy types, the evidence clearly indicates that CAC and MBI policies can have significant environmental benefits in developing countries. In addition to cataloging and reviewing this evidence, we discuss data and methodological challenges to augmenting it and suggest directions for future research. ”“The global South is home to the world’s worst environmental problems. For example, the 10 - cities with the world’s worst air quality are all in developing countries, as are many of the most ® polluted rivers (Marlier etal. 2016, WHO 2016). This pollution has serious adverse consequences fat both local and global levels. The World Health Organization (WHO) estimates that pollution Tes ‘causes 8.4 million deaths per year in developing countries, including 6.7 million from air pollution, " 0.8 million from water pollution and poor sanitation, and 1.0 million from toxic and industrial wastes (GAHP 2014, Landrigan & Fuller 2015). And each year, developing countries.Jed by China % and India, contribute roughly two-thitds of global greenhouse gas emissions (Eur. Comm. 2012). ° % e 9 ‘ G ¢ € ‘ ‘ { i Like their counterparts in industrialized countries, environmental regulators in developing countries rely principally on two types of instruments: command-and-control (CAC) policies, and toa lesser extent, market-based instruments (MBIs). CAC regulations such as emissions standards and technology standards mandate how much and/or how polluters abate. MBI regulations, by contrast, create economic incentives for abatement. The main types are emissions fees (fees per unit of emissions), marketable permits (emissions allowances that can be traded among polluters), and environmental taxes (taxes on inputs or outputs). A salient feature of CAC and most types of MBI policies is that they depend on environmental regulatory capacity to be effective: Regulators must check compliance and sanction violators, In developing countries, however, regulatory monitoring and enforcement are inhibited by gaps and inconsistencies in written legislation, weak (undermanned and underfunded) regulatory institu tions, a lack of political will for stringent enforcement, a shortage of public pollution control facilities ike wastewater treatment plants, and an abundance of difficult-to-monitor small and informal firms (Eskeland & Jimenez 1992, Russell & Vaughan 2003). As a result, the efficacy of CAC and MBI policies in developing countries is an empirical matter: We need rigorous evalua- tons to understand whether and under what conditions they generate the intended environmental benefits. ‘As discussed below, the past decade has seen a significant uptick in studies using rigorous empirical methods to evaluate the effects of CAC and MBI regulations in developing countries, To our knowledge, a comprehensive review of this evidence has yet to appear. This article aims to help fil that gap. To make the review both manageable and focused, we have used the following criteria for including studies. We focus on those that = are published in peer-reviewed journals; = examine policies in developing (versus transitioning) countries; = analyze “brown” regulations addressing air, water, solid waste, and hazardous wast ution, not “green” regulations governing natural resources such as re agricultur = analyze policies that require environmental regulators to check co violators; use rigorous impact evaluation methods that aim to control for cont address the efficacy of regulations in generating benefits. In addition, we put more emphasis on studies using longitudinal data the sectional data. “The remainder of the review is organized as follows. Section 2 model of a manufacturing plant's pollution abatement decision. The discussion of the literature. Section 3 briefly discusses data and evaluating environmental regulatory efficacy in developing regulation and includes subsections on stationary source regulations Blackman + Li + LinaM ‘The fundamental issue observed by the literature is that regulatory pressure is weak for the reasons listed in the introduction: inadequate regulatory institutions, a lack of political will for serine gent enforcement, a shortage of public pollution control facilities, and an abundance of diffeult- to-mnonitor small and informal firms. In terms of our graphieal model, the result is that the EMP. schedule lies close to the horizontal axis. Many of the studies reviewed in subsequent sections effectively cest whether imposition of @ CAC or MBI policy shifts that schedule up and shifts equilibrium emissions to the left 3. CHALLENGES IN EVALUATING ENVIRONMENTAL REGULATORY EFFICACY ‘Conducting rigorous empirical research on the efficacy of CAC and MBI regulation in developing countries is challenging for several reasons. First, credible data on environmental outcomes in particular, facility-level data collected by regulators—are scarce. In most developing countries, such daca simply do not exist: When they do, they are often self-reported and unverified. Also, they often suffer from selection bias: Plants that regularly self-report tend to be superior environmental pesformers. Given these problems, many of the studies we review in this article rely on other types of data, For example, among the 32 studies examining CAC regulations discussed below, only six rely on official plant-level data (Table 2). The rest use environmental quality daca at the level of anadministeative unit, monitoring station, or vehicle, or they use original plant-level survey data Second, disentangling the effects ofa specific regulatory policy on environmental performance from the effects of other regulatory policies ean be difficult. Government programs that address serious pollution problems are often multifaceted, entailing a suite of policies implemented con- currently. For example, China implemented a portfolio of policies to quickly imprave air quality in Beijing and other cities doring the 2008 summer Olympic Games, including closing major polluting facilities, mandating installation of abatement devices in coal-fired power plants, and ‘imposing more stringent emissions standards (Hle etal. 2016). And in Santiago, Chile, regulators rely on CAC emissions standards, tradable permits, license-based driving restrictions, and episodic plant closures to improve air quality (Mullins & Bharadwaj 2012). ‘Third, disentangling the effects of a specific regulation on environmental performance from the effects of omitted variables can be challenging. One reason is that regulators and communities surrounding dirty plants may pressure those plants to improve their environmental performance, a phenomenon that occursall over the world but is arguably particularly prominent in developing countries, where community pressure sometimes drives variation in environmental regulatory stringency (Blackman & Bannister 1998, Pargal & Wheeler 1996). A study that failed to control for community pressure might overstate the effect of regulation, Finally reverse causality cn confound treatment effects estimates, For example, regul ten target particularly dirty plants for enforcement (eg., Blackman etal. 2010). Ast of enforcement on environmental outcomes that did not control for that targeting c 4 spurious negative correlation between enforcement and environmental 4, COMMAND-AND-CONTROL REGULATION Overall, the emerging evidence on the effectiveness of CAC. indicates thatit can generate intended benefits; of the 32 studies: that CAC regulation has had some benefit (Tables 1 and 2), the evidence on regulating stationary source pollution Blackman Li + Linp98 Poe 780 purwap ‘ern aus fq SB soger pee doo | _spsepueas 382ey>51C. GO Fons EPS soon ound ee, ooo | _seau es ‘yes0us poe 1a rausunouauy sep aposida pevuuonau nes eto me = ep sa|_ sins |_| panne ead aa Sint RT msoaor eae =. (romney) ae, 386 Blackman 6 Li « LinIte2 Stadion the eee af er econo (CAC) and sat kerbasedinssent (MRD ensrommeneal Number Number_|_ with benefit Countries (number of studies) Seudy ope data gpe “Command-and-controk: stationary s00 = z a inh China @) ~Gross-sectional 7 s Mexico (2), Brazil (2), China (1), a5; iE [hie G), China G), Mexico ae 2 2 China 2) mal too o China (1), Mexico (1), 2 7 ‘Command-and-control: mobile sources Dany eae 10 % | Misco, China), Coloniis@,Eewador Qh Cie Oy “Miscellaneous 2 1 India (1), Mexico (1) 5 "Tes 2 7 Tl Cat @ 34] China 2), Meneo @, Chile, Inds G), real @), Coloatia Oy Bander (0) Haar bated iacnosan palsionovies Asean Panel i [etna mivaed Panel 2 Eero) ioaeetaal i a EERO) Cross-sectional r Tora 6 “Markecbased instrament: wadable permit Switch to natural gos Water ony Pare i ia = hin Panel 7 0 Chile ‘Cross-sectional 0 0 ‘Chile (1) Toral 2 o Total MBI ® 5 China (6), Chile @) ‘ype of regulation examined (covering both air and water pollution, only air, and only neat subsection summarizes the evidence on mobile source pollution, focusing on i popular license plate-based driving restrictions but also looking at other instrumen 4.1. Stationary Sources We found 20 studies tat focus on CAC regulations governing stat ‘on air and water pollution, seven on air pollution only, and four on 390° Blackman» Lio Linself-reported strength of CAC regulations helps explain plants hon emissions from Chinese power 4.1-2Ale potion. Seven silos oo sel tty ete a ea ulations. All but one find that these regulations have had some benefits. ae air pollution 4.1.2.1. Environmental episode days in Santiago, Chile. \n 1997, Chilean regulators establishes sources in Santiago on days when coarse particulate air pollution (PMO) is projected to exceed defined thresholds. These controls include shutting down major stationary sources, prohibitin the burning of biomass for home heating, and including adtinal vehicles in the Gye ding restrictions program (Gee Section 4.2.1). Mullins & Bharadwaj (2012) examine the regulations’ ef- fect on PMI0 air pollution and all-cause mortality. They use day-level data along with a matched difference-in-differences approach, essentially comparing changes in outcomes on days following declaration of an environmental episode with changes in outcomes on similar days with similar pollution levels prior to the implementation of the 1997 regulation. They find thar the regula- tion reduced PM10 by approximately one-fifth on the day of implementation and eut mortality among the elderly. Both effects persisted on days following the environmental episode. Troncoso etal. (2012) use similar daca along with a simpler event study approach to examine the effects of environmental episode days on a broader set of air pollutants, including carbon monoxide, nitro- gen oxides, sulfur dioxide, and fine particulates (PM2.5). They find that the regulation cut carbon monoxide, nitrogen oxides (both emitced mainly by vehicles), and both coarse and fine particulates: bur nor sulfur dioxides (mainly emitted by point sources). Finally, Coria (2009) uses stationary source-level panel data along with a duration model to determine whether environmental episode days (as well as Santiago’s emissions permit trading system) help explain these sources’ decisions to switch en masse from relatively dirty fuels to natural gas in the late 1990s and early 2000s. She concludes that they did not. Rather, the falling price of natural gas was mostly responsible for the switch. 4.1.2.2. The 2008 Beijing Olympic Games. Two studies examine the effects on air quality and health outcomes of the suite of CAC regulations China put in place to quickly improve air quality in Beijing, neighboring cities, and other host cities during the 2008 summer Olympic Games. Im- plemented between late 2007 and late 2008, these regulations affected both stationary and mobile sources. The stationary source measures included closing (either permanently or temporarily) major polluting facilities, relocating others, halting large construction projects, mandating in- stallation of abatement devices in coal-fired power plants, imposing more stringent emissions standards, and requiring oil and gas collection and recovery systems to be installed in gas stations, oil storage facilities, and tankers. Mobile source measures included increasing gas prices, ratchet- ing up mobile source emissions standards, restricting driving (see Section 4.2.1), and heavy-emitting vehicles. He et al. (2016) analyze the effect of all of these regulations data on PMIO generated by ground monitoring stations in 34 cities (8 trol cities) along with generalized difference-in-differences models. They4.2. Mobile Sources dics that focus on CAC regulations governing mole sources, including 10 on license plate-bosed driving restrictions, and two on other types of reguladon 7 42:1. License plate-based driving restrictions. License plate-baed driving restrictions santo reduce sir pollution and wallic congestion by prohibiting diver from wing thet vehicles ones more days a week based on the last digit ofthe license plate. Such policies have been implemented orldwide, including Beijing, Delhi, Bogocs, aulo, and Santiago. Recently, these regulations have recewed considerable atenion from economists, with 10 published studies. Findings have been decidedly mixed: Two studies have concluded that these programs have a it least some perverse effects, mo find no effects, and sx find significant benefits in at least a dozen developing country megacities w Quito, Medellin, San José, So P- 42.1.1. Mexie City. Menico's City's 1989 Day Without Driving (Hoy no Citeula, ENC) pro- sam is probably he work's best known and most stadied license plate-based driving restrictions initiative, Unel the rules were changed in 2016 to exempt all vehicles that could pass emissions tests, the program prohibited vehicles more than eight years old from being driven one week. day per week and one Saturday per month both in the Federal District and in 19 neighboring Counties from 5 Ast to 10 PM. Eskeland & Feyziough (1997) evaluate HNC's effeets om fuel com sumption and vehicle ownership. They use historical data to predict counterfactual outcomes and then compare predicted and actual outcomes. They find that the program actully increased gaso- line consumption, most likely because drivers purchased additional vehicles often gar-guzaling, highly pollucing clunkers—so that they could drive on the restricted day. Almost a decade after that seminal study appeared, Davis (2008) revisited the issue using more ‘comprehensive data and temporal regression discontinuity (interrupted time series) models that aim to identify breaks in trends at the time the program was first implemented. Outcomes analyzed were five types of air pollution (carbon monoxide, nitrogen dioxide, ozone, nitrogen oxides, and sulfur dioxide), gasoline sales, vehicle registration and sales, bus ridership, and taxi prices. Davis finds that, in general, ar pollution did not decline as result of HNC and actually increased during ‘weekdays. In addition, vehicle registration and sales increased and bus ridership decreased. Gallego et al. 2013b) use a regression discontinuity approach similar to Davis's (2008) but analyze only carbon monoxide and use models with different specifications (cg, different lengths of the event window, diferent order polynomials). The results confirm that HNC did not reduce air pollution in the long run, However, the authors find that it did cut ar pollution in the first several months after implementation, before drivers had a chanee to fally adjust by, for example, buying additional cars. 42.1.2, Beijing. 1g’ license plate~based driving restrictions program was first implemented in 2008 as part of a collection of policies aimed at improving air quality ahead of the Beijing summer Olympic Games (see Section 4.1.2). For its first two-month implementation period, which coincided with the Games, the program prohibited driving every other day seven days 3 week for all but three hours a day (midnight to 3 Am). Afterward, it prohibited driving one day per week, first between 6 Ast and 9 Pat and later (2009) from 7 Av to 8 pu. Evaluating the effects of Beijing’s driving restrictions program is challenging because it coincided with other air pollution control measures and beeause the one-day--week program was preceded by an every= “other-day program. Sun et al. (2014) examine the one-day-a-week program's effects on traffic speed and inhalable particulates. ‘They use day-level data along with a novel Blackmon » Li © Livdifference-in-differenc plates ending a epcneetiderifcation strategy based ona plausibly exogenous absence of license Strategy allows thera se eeeibmable to a cultural aversion to that number: They cake a measures. They hind see setanele the program's effects from those of concarrens ag quality speeds but had litle eas crvind Pestrctions suceeded in increasing average ciy-wide watt, Viard & Fu an eet on concentrations ofinhalable particulate mater (0f coarse paricvngs oa the one-day-a-eck program's effects on sn sit pollution index viewership, Thole gen OBe% oxides, and rulfr dioxide) and labor apply, proved by television eee ue Srey for disentangling the effects of the Program on air quality from those eee a son ple isto a advantage of not only the temporal variation in the stations co ronde Thee ation in its effec related to the proximiyy ofa quality monte, one area ‘They find that the program reduced air pollution byapproximately one-fifth and the program has bese nb 917% for workers with “discretionary workime.""They posiehat Vehiclec co ras been more effective than the HNC in Mexico City beeawse purchasing additional mena ynvent driving restrictions is more costly in Beijing, vehicles in Beijing are newer Provide evidence that rates of compliance with Bejing’: rogram are high. Finally, Lu 2016) examines the effects of changes in the rules of Beijing's one-day-a-week so0e a arer it was first implemented in September 2009. Specifally, he exsevines ay April 9 decision to narrow by two hours the dally window dting which coviedare cen force and J Bessie 2011 decision to, among other things, increase the maximum fine for violations Tle Dog Ry Zit pollution index data along with a regression discontinuity design similar to Davis’ (2008). Fe finds thatthe 2009 weakening ofthe policy led ta inenenenr pal coatana meet strengthening led to reductions, 42.1.3. Bogota. First implemented in 1998, Bogoti driving restrictions policy has undergone three significant revisions. In its initial incarnation, it restricted driving private vehicles for four hours a day: two hours during the morning rush hour and two hours. uring the evening rush hour. In 2001, the evening restriction was extended to cover public vehicles. In2004, the number ofhoure during which restrictions were in place for private vehicles was extended to three hours in both the ‘morning and evening. Finally, in 2009, the number of hours for private vehicles was extended to 14 hours per day, starting at 6 am. Bonilla (2016) uses hourly data along with a regression discon. inuity approach similar to Davis's (2008) to analyze the effects of the first two phases (1998-2004) and last two phases (2004-2009) of the Bogoté program on concentrations of carbon monoxide (a pollutant primarily emitted by mobile sources), gasoline consumption, and vehicle sales and registrations. He is unable to discern an effect of either set of phases on air quality or vehicle use. However, he finds some evidence that the latter two phases of the program actually led toa perverse increase in fuel consumption, vehicle ownership, and air pollution in morning peak hours. Zhang et al. (2017) use a similar empirical design to examine the effects of all four phases of the Bogoté program on seven air pollutants (carbon monoxide, PMO, three types of nitrogen oxides, ozone, and sulfur dioxide). They find that, in general, the program’s different versions have not led to significant overall improvements in air quality. More specifically, they find the program’s effects vary, increasing some pollutants and reducing others, partly as a result atmospheric chemistry. hourly data along witha triple differences design that takes a variation in the policy (which applies to only part of the city). reduced carbon monoxide pollution by 9-11% during peak eecaenahrierrg ©396 success is attributable to vigorous enforcement, the availability of tars and public transportation, the high cost of purchasing additional vehicles, and uncertainty about whether the PyP propane will be continued, which makes drivers reluctant to invest heavily in adapting to it Finally, de Grange & Troncoso (2011) examine the effecs of Santiago's driving restrietion program. Since the late 1980s, this program has prohibited diving vehicles without catalytic converters one day a week during morning rush hours in the auttimn and winter. Since the smid-19%s, a. more stringent version of the program was implemented: It applies to all vehicles ‘on environmental episode days, when air pollution is predicted to exceed certain thresholds (ee Section 4.1.2), De Grange & Troncoso use hourly data along with regression discontinuity models coexamine the effects of the original and additional programs on automobile, metro, and bus trips. They find that the original restrictions have not affected the use of private vehicles. However, the additional restrictions have reduced private vehicle use by 6% and increased metro use by 3%. 4.2.2. Other regulations on vehicular emissions. Vehicular emissions standards are common in developing countries. Oliva (2015) examines testing for compliance with such standards in Mexico Cir, focusing on a common corrupt practice: bribing testing center technicians to use emissions results from clean “donor ears.” She identifies test centers that engage inthis practice from telltale serial correlation in their emissions test results when they use the same vehicle for consecutive tests, Then, using data from “clean” test centers only, she estimates the fair probability of vehicles passing the test conditional on their observed characteristics, Finally, she estimates a structural model of car owners’ testing decisions. Her results indicate that roughly 10% of car owners cheat on emissions tests and that eliminating such cheating would significantly reduce vehicular ‘emissions. Although the study does not directly test for it, it clearly bears on the effectiveness of vehicular emissions standards. Finally, Kathuria (2002) examines a suite of more than 20 CAC regulations aimed at reducing vehicular emissionsin Delhi, inchuding conversion ofall commercial vehicles to compressed narural gs, phasing out older commercial vehicles, and ratcheting up emissionsstandards. The regulations ‘were implemented between 1998 and 2001. To identify the effec ofthese policies on particulate air pollution and nitrogen oxides, he uses daily 1999-2001 air quality data along with simple event study regression models. He is unable to discern an effect of any of these policies, a finding he attributes to the rapid growth in new vehicle registrations during the study period. 5. MARKET-BASED INSTRUMENTS, (ur discussion of market-based instruments focuses on emissions fees and tradable permits. Both require environmental regulators to check compliance and sanction violators, and thus their ef ficacy in developing countries is open to question. We leave aside environmental taxes (taxes on. inputs and outputs associated with pollution), which in general do not require such regulatory inputs. In theory, both emissions fees and tradable permits should have an una in reducing emissions. However, in developing countries, the actual effects are unch of tax evasion (Liu 2014), the informal economy (Bento et al. 2017), and as just monitoring and enforcement (Blackman 2009, Blackman & Harrington 2000). Although a number of articles examine various aspects of MBIs in developing ing their advantages and disadvantages relative to CAC (Blackman & Harringt studies focus on emissions fees in China and tradable permits in mixed (Table 2). Although five of the six studies of Chinese neither of the two studies of tradable permits in Santiago find Blackoan © Ls © Li5-1. China's Pollution Levy System ‘We found six studies ‘hat examine Chin Meiosis ‘China's pollution levy system, che largest Set of studies Focusing 5.1.1, Overview been extensively stu bop prin Sd udied. Rolled out nationally in 1982 aft sot ee natn, Se oc cnices29'oeepltnaeto at eee eee ‘ame, and noite. The formulas wed elclate lees be eed re.’ ‘There have been at least thr see se important changes, First, before 1993, levis were «d CAG emissions standards; since 199, all disch . all discharges 003, polluters were required to pay only for the pollutant ince 2003, polluters have been required ro pay for the three est total levy. Finally, before 1995, much of the levies could be rebated investments in pollution reduction equipment; sinee then, rebates for fave been converted to loans. A few researchers have con ened abetement incer plants pay levies onl to abate other pollu assessed only for disehs charges that exceede have been regulated. Second, before 2 Associated with the greatest total levy; si pollutants with the great 0 polluters to subsidize abatement investments h ‘cluded thar the peculiarities ofthe levy system’s design have damp. oot, Rar exampl, Lin 2013) argues that eease under the currents, lyon the three pollutants with the greatest total levy, they have no incentives tants. He also observes that for water pollutants, marginal levy rates are re duced above cersinthresholdn, which dines abate ncn soos to ce Pi adiition, Blackman & Harringron 2000) point ove that, becuse ofthe wide airude Chinese firms have in accounting, they may get back some of their levy payments through rebates, whesher (oF not they have invested in pollution reduction equipment. The net effect is to rechce levy rates, again weakening the system's abatement incentives, These design issues, slong with the monitor ing and enforcement constraints highlighted in the Introduction, imply that rigorous evaluations are needed to assess whether China's pollution levy system actually has environmental benefits, Probably the greatest challenge in conducting such evaluations is finding reliable data. In China, plant-level environmental performance data are only sporadically available and are not published by the government. Many published articles use plant-leve! data obtained from national or local environmental protection bureaus (EPBs). Asa result, the geographic and temporal scope of these studies is limited. Moreover, the levy system has rclied on self-reporting: Polluters eeport their emissions, and regulators inspect and verify the reports. The accuracy of self-reports is questionable. 5.1.2. Chinese plants’ responses to regulatory stringency. Research on the efficacy of the Chinese pollution levy system typically exploits variation inthe stringeney ofthe levy system actos plants. There are two main sources of such variation (Wang & Wheeler 2005). First, emissions standards are set jointly by the central and provincial governments. As a result, two otherwise identical plants located in different provinces can be subject to different levies. Second, inspection and levy collection are largely decentralized and the responsibility of local EPBs. Therefore, even aside from the heterogeneity just noted, the number of inspections and levy rates can vary significantly across plants. ‘A common methodological challenge in assessing the effectiveness of Chinese pollution levies is the codetermination of regulatory stringency and plant behavior: Plants change their environ mental performance as a result of increased regulatory stringency, but policy makers may abo See Lin 2013) and Wang & Whecler (2005) for dealed formulas se anmalreicaorg + Excinmmesta Regan in Deeg ona 397adapt the stringency of regulations (levy rates, inspections) to the environmental perform: tions approach (Jiang & McKibbin 2002, Wang & Wheeler 2005, Wang etal 2003), and oshere have used instrumental variables to tease out an exogenous portion of variation (Dasgupta et ah 2001, Jin & Lin 2014, Lin 2013, Wang 2002), Most studies assessing the Chinese levy system find that higher levy rates lead to improvements in environmental performance. Using eross-seetional data on 891 of China's top water polluters in 1993, Wang (2002) finds that investments in pollution abatement facilities and expenditures con operating those facilities responded strongly to effective levy rates, defined asthe levy acral pid for pollution per unit of emissions. He instruments forthe effective levy rate using region. level socioeconomic ancl demographic variables, such as income, education, population density, and environmental complaints. However, its not clea that these region-level variables do not directly affect plant-level investments and operating expenditures. Using 1993 cross-sectional state EPB data on 3,000 large air and water polluters, Wang & Wheeler (2005) find that water pollution emissions declined by 1% and air pollution emissions by 0.65% for each 1% inerease in the effective levy rate, suggesting that pollution levies generate a strong deterrent effect. The number of regulators per plant and complaints per capita are used we insrment Forth Geo ty Se ee affect outcomes. Jiang & McKibbin (2002) develop a structural model of equilibrium pollution using province- level pane] data on emissions of waste water, waste gas, and solid waste between 1992 and 1996. They find that pollution intensity was highly responsive to the levy rate. Jin & Lin (2014) examine the relationships between the effective levy rate, technological effi- ciency, and pollution intensity using province-level panel data from 1992 to 2008. They find that the effective levy rate significantly reduced pollution intensity. Using panel data on individual industrial polluters in Zhenjiang between 1993 and 1997, Das- gupta eral. (2001) analyze the effect on total suspended solids and COD water pollution and total suspended particulates air pollution of both pollution levies and regulatory inspections. Time tends are used to instrument for the number of inspections, a method that relies on the assump- tion that factors affecting emissions are uncorrelated with time. ‘The authors find that, although the number of inspections reduced emissions, the pollution levy had a statistically insignificant effect. Finally, Lin (2013) uses quarterly plant-level panel data froin the food, chemical, paper, and medicine sectors in an unidentified municipality to analyze how regulatory inspections aimed ar enforcing pollution levies affected COD water pollution. He uses plausibly exogenous variation in rainfall to instrament for the number of inspections while carefully distinguishing between actual emissions and those self-reported by plants under the fee discharge system. He finds that inspec- tions caused an increase in self-reported emissions. Acknowledging the institutional uncertainties outlined above, Lin does nit come to a conclusion about the effect on real emissions. His results are opposite those of Dasgupta etal. (2001), contrast that points to rich opportunities fo research, 5.2. Santiago’s Tradable Air Pollution Permits Program — Established in 1997, Santiago’sair pollution permit trading program initially matter. Later, its scope was expanded to include nitrous oxides and other program was implemented, ar emissions dropped: Platman + Li» Lin“ EGE MOBI PRADEEP XEROX CENTER HINDU COLLEGE & K.M.COLL enn Sango nd no dcr bene had ny eng ones ery eee problems they simed ra solve. In contrast, the programs in Beiing and Quine cere stsetions policies can have significant benefit, they donot necessarily mph hat pose are cost-effective. Emerging evidence on the costs of din baesceaee "7 significant cont on drives (Blackman etal. 28a; Calls ar segs en ‘impose Aside from the evidence on driving restrictions and Santiago's episodic CAC rollin levies, sw ues ges hae pollution regulations, Se implemented on severe ai pollation benefits. Finally, ewo studies suggest tha ete Beings ers to mprve ai quali aor the 2008 Olympic Games, using a suite of CAC policies, were effective, . : 3 6.3. Causal Mechanisms The evidence we have reviewed provides some hints about causal mechanisms that drive effee- tiveness of CAC and MBI reglations. Again, we have the most evidence on cease plate-based Ariving restrictions. Although 10 studies are too small a sample from which wo dese hablo cies ncesaboue causal mechanisms, the authors who documented sucessful program i Being sd Quito have speculated about the factors correlated with effectiveness (Carrillo ee al 2016, Vsed & Fu 2015), These include stringent program enforcement, the high cost and limited avaabi- ity of vehicles, the low cost and high quality of publi transportation, and in the ease of Quivo, uncertainty about the longevity of restrictions, Two studs suggest that in developing counties corruption plays a major role in blunting the eect of CAC regulation, and that interventions simed at checking corruption, such as auditing the auditors, may enhance efficacy (Duflo etal. 2013, Oliva 2015), Finally, two studies hine that ying administrative promotions and job security to regulatory enforcement can enhance efficacy (Jin & Lin 2014, Kahn et al.2015). 7. FUTURE RESEARCH Although rigorous quantitative evaluations of CAC and MBI regulations in developing countries are being published more and more frequently—an encouraging trend—the evidence base remains (quite thin overall, Several gaps need to be filled, Some gaps correspond to the isues discussed above, First, studies are needed on countries other than upper-middle-income and high-income countries such as China, Mexico, and Chile, Although data availability is a limiting factor, there surely are middle- and low-income countries where eve pethaps using remotely sensed data on air and water quality (Donaldson & Storeygard 201 Second, even for the countries where we have evidence, such as China, Mexico, and Chile, more studies of underrepresented regions are needed. For example, most of the existing evidence ton China's pollution levy system is based on data provided by municipalities in China's eastern provinces. Studies on other parts of China would paint a more complete picture ofthe system. Third, further evidence is needed on MBI regulations in developing countries. We er find only eight rigorous studies that met our criteria, and these studies focus on only two Blackmon + Li + Lin a araPelicies—pollution levies in ‘China is certainly mt evidence bse ar whole does no enabl countries. Given thelinited sum ee-bated factors from th sheet cay of Minder t0 dise eee dsentangle the een needed on specife ype uf CAC regulations f ly ate plate-based driving restriction: ere » outcome variables such as health. ep aged interest tha emissions. S missions. Sixth, research : alternative iden ‘tive ms 285 of studies on other polices, Fifth beofgreater policy such a regres ih or propensity sore matching, migh help eseacher ; nce tron Chines pln Seen feel has tc mechs secede CAE aN pos ota Soe costs of CAC and MBI policies in dev i ib ts rents have suggested that MBI policies ate likely to, i te rvonmens odin ay eropng oes il Rus 1 Ban go 2000), Additional evidence en the costs af CAC and MBI policy as well ason tr fay, needed to advance this deat DISCLOSURE STATEMENT The authors are not aware of any affiliations, memberships, funding, or financial holdings thi might be perceived as affecting the objectivity ofthis review ACKNOWLEDGMENTS We are grateful to Sally Atwater for editorial asisance and an anonymous reviewer for help comments and suggestions. LITERATURE CITED 1G, Rowell C2002 Favronmenta poy for dvelping cunts ma Se Tec 1863-70 aa, Li 201, armen pay prorer f enfoml aw: Fr Dron Sch, Appl Econ, Manag, Corell Univ Iuhach NY : in 200, Clon dakar fo propa: Incr fer peso egde=nt 7. = Manag, 90(1)101-19 Blaclnan A. Alpiar Fy Carin Dy without Driving prog pressure and clean techooo iy raiinal Mexican briehmakers 7B DYancer MR, 2018. A eningent vain appro eiating Tavoy costs: Mexico's Jae on. Boe OTS Blackman A, Bannister C) ae cern co arvie analysis of the adoption of propane 1998, Commaniy ae ives in developing counties: lessons fromm Blakes A, Harn ind pin JE De = nt un crn mm sa on Ply Sod 10 internavonal ia Blackman A, Kildegoard & jean leather tanneries atc + ExroesR 2424 HINDU COLLEGE & K.M.COLLEGE MOBILE: 813046: ROX CENTER pRADEEP x