cities, industrialization, and job creation: evidence from … · 2019-07-29 · cities,...

25
Cities, Industrialization, and Job Creation: Evidence from Emerging Economies * Thor Berger Chinchih Chen Carl Benedikt Frey June 13, 2017 Abstract In this paper, we estimate multipliers of jobs in the tradable sector on non-tradable employ- ment in cities in Brazil, China, India, Mexico, and South Africa. Building on an emerging liter- ature on local labour multipliers, we show that average multipliers in emerging economies are smaller relative to existing estimates for the United States, while the multiplier for skilled jobs is 6 to 9 times larger both according to simple OLS estimates and when using an instrumental variable strategy that exploit nation-wide changes in employment and initial cross-city industry specialization. While there is growing concern that the manufacturing sector is not absorbing as many workers as it once did, we document that its indirect impact on service employment is substantial. We further conclude that emerging economies can generate significantly more jobs in the nontradable sector by shifting towards more skill-intensive production. JEL: J23, R11, R12, R23 Keywords: Cities, manufacturing, multipliers, emerging economies * Berger: Department of Economic History, School of Economics and Management, Lund University & Oxford Martin School, University of Oxford. (E-mail: [email protected]) Chen: Oxford Martin School, University of Oxford. (E-mail: [email protected]) Frey: Oxford Martin School, University of Oxford. (E-mail: [email protected]). 1

Upload: others

Post on 24-Mar-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Cities, Industrialization, and Job Creation: Evidence from

Emerging Economies∗

Thor Berger Chinchih Chen Carl Benedikt Frey

June 13, 2017

Abstract

In this paper, we estimate multipliers of jobs in the tradable sector on non-tradable employ-

ment in cities in Brazil, China, India, Mexico, and South Africa. Building on an emerging liter-

ature on local labour multipliers, we show that average multipliers in emerging economies are

smaller relative to existing estimates for the United States, while the multiplier for skilled jobs

is 6 to 9 times larger both according to simple OLS estimates and when using an instrumental

variable strategy that exploit nation-wide changes in employment and initial cross-city industry

specialization. While there is growing concern that the manufacturing sector is not absorbing

as many workers as it once did, we document that its indirect impact on service employment is

substantial. We further conclude that emerging economies can generate significantly more jobs

in the nontradable sector by shifting towards more skill-intensive production.

JEL: J23, R11, R12, R23Keywords: Cities, manufacturing, multipliers, emerging economies

∗Berger: Department of Economic History, School of Economics and Management, Lund University & OxfordMartin School, University of Oxford. (E-mail: [email protected]) Chen: Oxford Martin School, University ofOxford. (E-mail: [email protected]) Frey: Oxford Martin School, University of Oxford. (E-mail:[email protected]).

1

Page 2: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

1 Introduction

In a seminal review of urban development, Paul Bairoch (1988, p.286) asserts that “[i]ndustrializationfosters urbanization, and cities promote further industrialization, engendering in the process a grow-ing number of specifically urban forms of employment: jobs for street car drivers, meter readers forwater, gas and electrical systems [...] and so on.” Since the Industrial Revolution, this process hasconstituted a cornerstone of most countries’ development strategy, inducing workers to shift fromrural areas and argicultural activities to cities, where they engage in the production of manufacturedgoods and services (Michaels et al., 2013; Gollin et al., 2002). Nevertheless, the secular declinein countries’ peak manufacturing employment share over the course of the twentieth century hasrecently raised the concern that industrialization has ceased to be an engine of job creation, aslabour-saving technology put emerging economies at risk of “premature deindustrialization” (Ro-drik, 2015).

The contribution of industrialization to job creation does however not only stem from its di-rect employment impact: one additional manufacturing job in any given city generates more localdemand for nontradable goods and services, leading to an expansion of urban employment (Park,1989). In this paper, we estimate long-term employment multipliers of jobs in the tradable sectoron non-tradable jobs in some of today’s largest emerging economies: Brazil, China, India, Mex-ico and South Africa. Our approach builds on recent work by Moretti (2010) showing that eachmanufacturing job in the United States has created 1.6 jobs in the local nontradable sector and thatadding one additional skilled job in the tradable sector generates 2.5 jobs in a given city.1 Foremerging economies, the magnitude of these multipliers have important implications for regionaleconomic development policies. Assuming that the objective of local economic development poli-cies is to maximize local employment, for example, it is important to know where subsidies shouldbe directed, as the multiplier is likely to vary across industries and skill groups.2

The magnitude of the multiplier effect depends on four factors (Moretti, 2010; Moretti andThulin, 2013; Moretti, 2011). First, stronger consumer preferences for nontradables implies ahigher multiplier as more is spent on local goods and services. Second, the use of labour-savingtechnology in the nontradable sector reduces the size of the multiplier, all else being equal. Third,the magnitiude of the multiplier hinges upon the type of jobs being created in the tradable sector:skilled workers typically exhibit higher earnings and therefore generate more additional demand fornontradables. Fourth, there are offsetting general equilibrium effects on wages and prices, which

1These new jobs are split between existing residents and new residents who move from elsewhere, depending onthe degree of geographical mobility: in countries undergoing more rapid urbanization the influx of workers from ruralareas is likely to account for a higher share of new jobs relative to existing residents.

2Place-based economic policies, for example, are widespread in the United States. Indeed, it is rare for a largeproduction or research facility to open in the U.S. without the provision of some form of subsidy from the relevantlocal government (Greenstone and Moretti, 2004; Greenstone et al., 2010).

2

Page 3: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

are determined by the elasticities of local labor and housing supply; an increase in labour costsreduces the supply of nontradables and cities with more constrained housing supply will experiencelarger increases in costs of living, in turn reducing the size of the multiplier (Saks, 2008).

In our empirical analysis, we use data on 93 million workers located in cities in Brazil, China,India, Mexico, and South Africa from the Integrated Public Use Microdata Series International(IPUMS-I) to quantify the impact of job creation in a city’s tradable sector on nontradable em-ployment. Our baseline estimations compare changes in skilled and unskilled tradable jobs andnontradable employment across cities, allowing for endogenous adjustment of prices and factorreallocation. OLS estimates show that local multipliers range from 0.5 in Brazil to 6.3 in Mexico,while the multiplier effect for one additional skilled worker is typicallly at least three times larger.Using various definitions of skilled jobs, we show that one additional worker with a secondarydegree in the tradable sector creates between 2.9 (South Africa) and 29.9 (Mexico) new jobs inthe local nontradable sector, while the multiplier effect for an additional worker with a college de-gree is substantially higher, ranging from 16.1 in South Africa to 21.1 in India.3 Although theseestimates are robust to using various definitions of cities as well as to including a wide range ofcity-level controls, including demographics and unemployment rates, it is still possible that thereare unobserved local labor market shocks that may confound our OLS estimates. To account fortime-varying labor supply shocks, we use an instrumental variable (IV) technique that isolates ar-guably exogenous shocks to the local labor demand in the traded sector by exploiting nation-widechanges in industry-level employment interacted with initial cross-city specialization across indus-tries. IV estimates reaffirms the main findings of the OLS analysis, which suggests a causal impactof job creation in the tradable sector on nontradable employment.

Our study relates to two literatures. First, our paper is most closely related to an emerging lit-erature on local labour multipliers (Moretti, 2011, 2010; Moretti and Thulin, 2013; Moretti andWilson, 2014; Marchand, 2012; Faggio and Overman, 2014; Van Dijk, 2014).4 For example,Moretti (2010) finds that manufacturing jobs in the United States have significant spillovers onlocal demand.5 In addition, Moretti and Thulin (2013) document differences in the magnitudeof local multipliers between Sweden and the United States, arguing that relatively low multipli-ers in Sweden stem from a more compressed wage distribution and less elastic labour supply, due

3While the multiplier of 34.7 in China is the largest, this result is not statistically significant. Similarly, althoughboth China and Mexico exhibit larger college multipliers, IV estimates are statistically insignificant.

4A number studies also use variation at the micro-level to estimate the effect of government stimulus programs.of A recent study by Wilson (2012) uses cross-state variation in fiscal stimulus outlays to estimate the jobs multiplier.Furthermore, Mian and Sufi (2010) examine examine the ability of the government to increase consumption by evaluat-ing the impact of the 2009 “Cash for Clunkers” program on short and medium run auto purchases, exploiting variationacross U.S. cities.

5Marchand (2012) further examines the impact of energy extraction jobs on the local nontradable sector, findingthat energy extraction jobs created during a boom period leads to the creation of approximately three construction jobs,two retail jobs, and four and a half service job.

3

Page 4: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

to more generous unemployment benefits and public subsidies. Relative to previous estimates foradvanced economies such as the United States and Sweden, we document that the multiplier in gen-eral, and the skilled multiplier in particular, is substantially higher across all investigated emergingeconomies.6 Our findings thus speak to the intuition that the skilled multiplier depends on differ-ences in wages between skilled and unskilled jobs (Moretti and Thulin, 2013), and that returns toeducation are typically higher in emerging economies relative to advanced ones (Psacharopoulosand Patrinos, 2004; Psacharopoulos, 1994), contributing to higher multipliers for skilled workers inemerging economies.7 In addition, the abundance of cheap labour implies less use of labour-savingtechnology in the nontradable sector (Hornbeck and Naidu, 2014), contributing towards a higheraverage multiplier. Finally, because the magnitude of the multiplier crucially depends on the elas-ticity of local labor supply, cross-country variation depends on labour mobility and labour marketinstitutions: emerging economies are often characteristed by migration from rural areas to cities,inducing higher geograpical mobility, and have less generous public subsidies and unemploymentbenefits, leading to a higher multiplier.8 We thus add to this literature providing the first evidenceon the size of labour multipliers in emerging economies, documenting that these are significantlyhigher relative to previous estimates for advanced economies.

Second, our study is related to a growing literature on the employment-generating capacityof the manufacturing sector in developing and emerging economies (Morawetz, 1974; Baer andSamuelson, 1981; Park, 1989; Rodrik, 2015; Amirapu and Subramanian, 2015). In particular, Ro-drik (2015) shows that the manufacturing share of employment steadily has declined among emerg-ing economies since the 1960s, as manufacturing processes have become increasingly automated.We add to this literature, showing (i) that manufacturing indirectly provides an important engineof job creation in emerging economies by generating additional demand for local nontradables;and (ii) that emerging economies can achieve more rapid job creation in the nontradable sectorfrom industrialization by shifting their export composition towards skill-intensive products, thatare associated with a larger multiplier.

The remainder of this paper is structured as follows. In section 2, we describe our data sourcesand empirical strategy. Section 3 discusses our key findings, and provides additional robustnesschecks. Finally, in section 4, we derive some conclusions and implications for policy.

6While the skilled multiplier is substantially higher in all investigated emerging economies relative to the UnitedStates, the average multiplier for the tradable sector taken as a whole is also larger in China, India and Mexico.

7In India, for example, growing wage inequality has been induced by skill-biased technological change (Kijima,2006).

8The multiplier should also be higher in countries that are characterized by higher rates of rural-urban migration,as migrants initially typically earn lower wages (Vijverberg and Zeager, 1994).

4

Page 5: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

2 Data and Empirical Framework

Our analysis takes its departure in the model outlined by Moretti (2010) in which a finite numberof local economies (“cities”) uses labor to produce a range of tradable and nontradable goods andservices. Prices for the set of tradable goods produced in each location are set in international ornational markets and are thus exogenously given for each city, while prices for nontradable goodsand services are set locally. In this framework, an increase in the number of jobs in the tradablesector has two distinct effects: (i) it affects employment in the tradable sector directly; and (ii) ithas indirect effects on the demand for nontradable goods and services. As the demand for non-tradable goods—such as restaurants or taxi fares—increases due to an expansion of employmentand income in the tradable sector, the creation of tradable jobs generates additional employment inthe nontradable sector through the local multiplier. As outlined in the introduction and as furtherdiscussed in Moretti (2010), Moretti (2011), and Moretti and Thulin (2013), the size of this localmultiplier is determined by four factors. First, stronger consumer preferences for nontradables im-plies a higher multiplier as a larger share of incomes are spent on local goods and services. Second,a more intensive use of labour-saving technology in the nontradable sector reduces the size of themultiplier. Third, the magnitude of the multiplier hinges upon the type of jobs being created inthe tradable sector: skilled workers typically exhibit higher earnings and therefore generate moreadditional demand for nontradables, which should result in a larger local multiplier for skilled rel-ative to unskilled jobs. Fourth, there are offsetting general equilibrium effects on wages and prices,which are jointly determined by the elasticities of local labor and housing supply. For example,an increase in labour costs reduces the supply of nontradable goods and services and cities with amore constrained housing supply will experience larger increases in costs of living, which in turnserves to reduce the size of the multiplier (Saks, 2008).

A simple starting point to estimate the local multiplier is to regress the change in the number ofworkers employed (4E) in the nontradable (N) sector on the change in employment in the tradable(T) sector between year t and t-1 across a sample of cities i:

4ENit = α +ϑt +δ4ET

it + εit (1)

where α is a constant, εit is a random error term, and ϑt is a period dummy, which captures shiftsin nontradable employment that affect all cities (i.e., the national economy) the same way. In thiscase, δ would capture the number of additional nontradable jobs that are created due to the additionof one tradable job. Our main empirical analysis focuses on the effects of changes in tradable jobson the nontradable sector, though additional estimates distinguish between employment in skilledand unskilled tradable sectors and also examine changes the impacts of growth in the tradable sectoron that sector itself.

5

Page 6: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Interpreting δ as the causal impact of job creation in the tradable sector, however, requires anumber of assumptions that are unlikely to hold. Most importantly, if a city-wide shock to em-ployment causes both the tradable and non-tradable sectors to expand, we would attribute part ofthis increase to the local multipliers of jobs created in the tradable sector. For example, if a cityintroduces changes to its tax policy that affects the labor supply, employment in both the tradableand nontradable sector may be affected. To identify the local multiplier δ we therefore ideally wantto isolate exogenous shifts in employment in the tradable sector that allow us to identify the causaleffects of job creation in the tradable sector.

Following a long line of analysis in regional economics, we exploit variation in national industrygrowth rates and initial cross-city industry specialization to obtain exogenous variation in employ-ment growth at the city-level as the basis for our IV analysis (e.g., Bartik, 1992). To fix ideas,consider two cities that are specialized in different tradable industries. An upswing in demandfor the output of one of these industries will cause the city that is disproportionately specializedin the production of that good to experience an increase in demand that is plausibly uncorrelatedwith city-level factors. For example, an increased international demand for textiles will lead toan increase in employment in cities that initially specialized in textile production that is plausiblyuncorrelated with contemporaneous city-level shocks. Following Moretti and Thulin (2013) our IVanalysis therefore uses the following instrument to predict employment in all tradable industries j

in city c in year t:

∑j

ETi jt−1(ln(E

Tjt −ET

i jt)− ln(ETjt−1−ET

i jt−1)) (2)

where ETi jt and ET

i jt−1 is the employment level in tradable industry j in city i in year t and t-1

while ETjt and ET

jt−1 is the national employment in industry j in year t and t-1 respectively. Thisinstrument thus isolates variation in employment growth in the tradable sector for each city that isdriven by changes in national employment growth, excluding the employment in city i itself, acrossall tradable industries. Crucially, these aggregate shifts are unlikely to be correlated with city-leveleconomic conditions thus plausibly yielding exogenous changes in tradable employment.

2.1 Data

For our analysis, we use data from the IPUMS-I, where the integrated census data are derived fromnational census surveys (Center, 2015). To limit the potential bias resulting from the differentialeconomic and technological development across countries, we focus on five of the largest emerg-ing economies—Brazil, China, India, Mexico and South Africa—for which IPUMS-I providesconsistently defined variables for census years between 1980 and 2010. Though most countries rundecennial census surveys over time, some have conducted interval surveys between decades. To

6

Page 7: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

keep a consistent time span, we construct the sample by selecting the decennial censuses between1980 and 2010 as well as those with a time span between two consecutive census surveys no morethan fifteen or less than five years apart.

Our empirical analysis sets out to estimate local multipliers for urban economies and we con-sequently limit our samples to include workers employed individuals between 15 and 65 years thatreside in urban areas. Since there is no urban status variable for China samples, we only selectthe observations in cities, which are closest to the definition of metropolitan areas (?). In addition,since there exist size differences in population and territory across these five countries, the mini-mum threshold of 20,000 of urban/city population is set to reduce the potential scale bias. We alsoreport results from different thresholds (100,000; 500,000 and 1 million) to test this potential scalebias, which yields qualitatively similar results.

To establish consistent local labour markets across countries over time, we use the second majoradministrative level of geography, recorded in IPUMS-I, which provides the most detailed spatialunit.9 To account for the fact that IPUMS-I has kept updating the consistent boundary unit at thislevel, we use the harmonised geographical unit “GEOLEV2” for Mexico and South Africa, whichindicates the second major administrative level of geography where an household resides. ForBrazil, the comparable municipality, “MUNIBR”, which is the second major administrative geo-graphical unit in Brazil integrated by IPUMS-I, is used. While there is no comparable second majoradministrative geographical unit for China and India, IPUMS-I have constructed name harmonisedunits, “GEO2 CNX” and “GEO2 INX”, for China and India, respectively. Both units have incor-porated the change in administrative status, such as the change in their identification as cities versusprefectures in China or the boundary change in higher administrative level of geography in India.10

Hence, IPUMS-I has considered these two units are generally comparable over time.

Using IPUMS-I’s general industry recode, we follow Moretti’s (2010) definition of tradable andnon-tradable sectors.11 We include manufacturing industries in the tradable sector, while the non-tradable sector comprises construction, electricity, gas and water and services. We further excludeagriculture, fishing, forestry and mining as these jobs are typically rural. In addition, the durable

9The second major administrative level of geography differs in terms of scale and terminology. IPUMS-I has at-tempted to construct spatio-temporally harmonized geographical unit to provide spatially consistent boundaries acrosssamples in each country.

10The following cities in 1990 were identified as prefectures in 1982: Changde, Chaoyang, Huayin, Jiaxing, Jinhua,Jining, Putian, Sanming, Shaoxing, Siping, Tai’an, Tianshui, Tieling, Tonghua, Weifang, Yangzhou, Yantai, Yueyang,Zhenjiang, and Zhoushan. IPUMS-I adjusts the status of these places as cities in 1982 to be comparable between 1982and 1990. In general, India’s regions are comparable across samples, However, Goa and Daman and Diu were onestate in 1983 and 1987, but they are identified separately in the later samples. GEO2 INX codes the following states tothe regions they were part of between 1983 and 1999: Chhattisgarh to Madhya Pradesh, Jharkhand to Bihar (Southernregion), and Uttaranchal to Uttar Pradesh (Himalayan region and part of the Western region).

11We use crosswalks to bridge each country’s codes to IPUMS-USA codes, and then aggregate into 2-digit ISICRev. 3 industry codes. Based on ISIC Rev. 3 codes, durable tradable sector include industry 15, 16, 17, 18, 19, 21, 22,23, 24 and 25; the non-durable tradable sector contain industry 20, 26, 27, 28, 29, 30, 31,32, 33, 34, 35 and 36.

7

Page 8: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

and nondurable tradable sectors are constructed according to the definition of IPUMS-USA 1990industry codes of manufacturing sector. Furthermore, while Moretti (2010) defines skilled workersas having obtained a college degree, we rely on two different definitions, as only a fraction of work-ers in emerging economies have a college education. Using the educational attainment variable inIPUMS-I, the first classification defines that the skilled jobs are those fulfilled by workers with sec-ondary degree completed and above, and unskilled jobs are those with primary degree completedor less. In a second definition, we further define skilled workers as the ones that have obtained acollege degree.

Figure 1 and 2 document the tradable and non-tradable share of total employemnt in urbanareas, showing that Brazil, China, India and Mexico all have experienced a decline in the shareof employment in the tradable sector, while the share employed in the non-tradable sector hasincreased. South Africa, on the other hand, has experienced an expansion of its tradable sector,while the employment share in the non-tradable sector has remained stable. Figure 2 further showsthe change of the share of skilled jobs over time, documenting that the supply of skilled workersin our sample has generally increased, the exceptions being college educated workers in China andworkers with a secondary college degree in South Africa.

Our final sample includes more than 93 million worker observations, comprising integratedcensus information on demography, education, and work; including age, employment status, sector,sex, and educational attainment. Table 1 provides descriptive statistics of the key variables ofinterest for each country.

[Figure 1 about here.]

[Figure 2 about here.]

[Table 1 about here.]

3 Empirical Estimates

In this section, we present our estimates of local multipliers in Brazil, China, India, Mexico, andSouth Africa. We begin by presenting OLS and IV estimates of the impact of job creation in thetradable sector on nontradable jobs and then proceed to examine changes in tradable-on-tradableemployment.

3.1 Local Multipliers in Emerging Economies

Table 2, columns 1 and 2 presents OLS estimates from equation (1) that correspond to the additionalnumber of jobs created in the nontradable sector due to an increase in employment in the tradable

8

Page 9: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

sector. Each row reports city-level regressions for Brazil, China, India, Mexico, and South Africadrawing on data for the 1980-2010 period. Column 1 reports changes in non-tradable employmentresulting from changes in the tradable sector conditioning only on time fixed effects, where theestimates suggest, for example, that an additional tradable job created in an Indian city on averageleads to the creation of 1.6 additional jobs locally in the nontradable sector. To address potentialomitted variable issues, the estimates in column 2 include controls for time-variant local labor mar-ket characteristics as suggested by Faggio and Overman (2014): the total labour force, the shareof female employment, and the share of employment in different age groups. Although estimatesare relatively stable for China, India, and South Africa, coefficients change quite substantially inthe case of Brazil and Mexico, which suggest that omitted city factors may be biasing the baselineresults in column 1. Moreover, the negative multipliers for Brazil are suggestive of a negative un-observed shock to employment, which further motivates the need to isolate exogenous changes intradable employment.12 While different time periods of investigation does mean that the reportedmultipliers are not directly comparable across countries, the estimates together suggest that multi-pliers in emerging economies are typically lower than existing estimates for advanced economiesas the average multiplier for all countries in our sample is lower than the 1.6 figure reported for theUnited States (Moretti, 2010).

As discussed above, local multipliers are expected to be higher for skilled jobs as the higherincomes of skilled workers typically generate more demand for nontradable goods and services.Consequently, we allow for a heterogenous impact for skilled and unskilled jobs by separating thetradable sector into skilled and unskilled jobs. Using two alternative definitions of skilled tradablejobs, columns 3 and 4 report estimates of the number of additional nontradable jobs that emergeas a result of one additional skilled job in the tradable sector. Column 3 examines the multiplierassociated with skilled manufacturing jobs defined as jobs held by workers with at least a secondarydegree. The multiplier for skilled jobs is typically at least 3 times larger than for the tradable sectoras a whole and substantially larger than estimates for the United States, with the exception ofChina. Alternatively, column 4 identifies skilled jobs as those in which workers have obtaineda college degree as in Moretti (2010). Estimated college multipliers range between 6 and 27,and are thus considerably larger than our estimates of the secondary degree multiplier, consistentwith studies documenting large skill premiums across emerging economies (Psacharopoulos andPatrinos, 2004; Psacharopoulos, 1994).13 Relative to previous estimates for the United States, wenote that the college multiplier is approximently 3 to 10 times larger in the five emerging economies

12As discussed more in detail below, IV estimates show that when using nationwide shifts in employment as a sourceof identification the estimated multiplier of tradable jobs on the nontradable sector is positive.

13Other studies also show that educated workers in emerging economies are more productive and rewarded thereafter(Jones, 2001).

9

Page 10: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

we examine.14

[Table 2 about here.]

However, a main empirical concern is that the OLS estimates are likely to be biased due tounobserved city-level factors that covary with changes in nontradable and tradable employment,which implies that the OLS estimates are not neccessarily informative about the actual size of themultiplier. To alleviate such concerns, we next proceed to examine the causal impact of tradablejobs on the non tradable sector by employing an IV strategy that use nationwide changes in industryemployment interacted with each each city’s initial industry structure to predict changes in tradableemployment. Although a large literature emphasizes that weak instruments may lead to biased IVestimates (Staiger and Stock, 1997; Stock et al., 2002; Stock and Yogo, 2005), these shift-shareinstruments are typically strong predictors of subsequent changes in tradable employment at thecity-level; the first stage Kleibergen-Paap F-statistics are typically large enough to allow us to rejecta maximum 10 percent IV bias thus largely reducing concerns that our IV estimates are biased dueto weak instruments.

Table 3, columns 1-3 report IV estimates from equation (1) where the effect of adding skilledmanufacturing jobs is allowed to differ from the effect of unskilled manufacturing jobs, again usingthe two alternative definitions of skilled jobs. Column 1 reports IV estimates that are typically largerthan their corresponding OLS estimates reported in Table 2, column 2, which suggests that unob-served negative shocks to nontradable employment are correlated with job growth in the tradablesector. As in the previous analysis, estimated local multipliers vary considerably across countriesranging from around 0.5 in Brazil to more than 6 in Mexico. Interestingly, with the exception ofSouth Africa, multipliers are consistently larger when instrumenting for city-level changes in trad-able employment with the shift-share instrument, which exploits nationwide employment changesand initial industry specialization to isolate plausibly exogenous shocks to tradable employmentfor individual cities. Columns 2 and 3 report IV estimates where the shift-share instrument is skill-specific: in the first stage we use nationwide changes in skilled tradable jobs interacted with eachcity’s initial share of skilled jobs. Using a version of the shift-share instrument that is skill-specific,we find that the elasticity is significantly larger for skilled workers. These findings lend furthersupport to the argument that skilled jobs pay higher earnings than unskilled ones and therefore gen-erate more demand for local goods and services: the multiplier for skilled jobs is still substantiallyhigher than for the tradable sector taken as a whole (column 1), although we note that estimates of

14A central empirical concern is that our results are biased due to different definitions of cities across countries. Toat least partly alliviate such concerns, Table 6 reports multipliers using various city definitions, exluding cities with apopulation of less than 100,000 or 500,000 or 1 million. Crucially, the reported multipliers only change slightly acrossdefinitions, with the exception of Mexico, where the skilled multiplier turns insignificant as the number of cities in thesample drops from 248 to 64 when only cities with a population of 500,000 or more are included.

10

Page 11: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

the college multiplier for China and Mexico are statistically insignificant, presumably due to thesmall share of college educated workers.

Taken together, results presented in this section suggest that local multipliers are substantiallylarger in most emerging economies when compared to estimates for the United States, and thatskilled jobs have more economically and statistically significant positive spillovers on local em-ployment in the nontradable sectors. In the next section, we examine the impact of job creation inthe tradable sector on the tradable sector itself.

[Table 3 about here.]

3.2 Additional Results: Tradables-on-Tradables

The way job creation in the tradable sector can fuel job growth in local nontradable industries, theexpansion of one tradable industry might also impact on employment in other tradable industries,although the direction of the effect is a priori unclear. On the one hand, an increase in production inone given industry increases local demand for intermediate goods and services (Park, 1989), and ifagglomeration economies exist an increase in production could result in more local clustering (see,for example, Greenstone et al., 2010). On the other hand, a local increase in labor costs reducesemployment across tradable industries as, unlike the case of nontradable goods, the price of tradablegoods is set on the national market and cannot adjust to local economic conditions (Moretti, 2010).Thus, ultimately, the size of these multipliers remain an empirical question.

To examine the impacts of job creation in the tradable industries on other tradable industries,we divide it into durable tradables and non-durable tradables.15 Table 4 and 5 presents our resultsfrom estimating the impact of one additional job in the durable tradable sector on employment inthe nondurable sector and vice-versa using both the baseline OLS and IV approach. Reassuringly,multipliers for the tradable sectors are consistently substantially smaller in magnitute relative toour estimates for the nontradable sector reported in Table 2, which is consistent with the theoreticalprediction that spillovers from the tradable sector is smaller on other tradable sectors relative tonontradable industries. Furthermore, multipliers from IV estimates reported in panels B of Table 4and 5 are typically smaller in size relative to our OLS estimates: for the nontradable sectors our IVestimates are substantially larger, implying that the causal impact on jobs in the tradable sector isrelatively small, which more broadly is suggestive of relatively weak industry linkages in emergingeconomies.

[Table 4 about here.]15An alternative approach is to randomly divide the tradable sector into two subsets of industries and estimate the

effects of job creation in the tradable sector on itself as in Moretti (2010). Using that alternative approach, however,yields qualitatively similar results that are available from the authors upon request.

11

Page 12: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

[Table 5 about here.]

4 Conclusions

As urbanization and industrialization has accelerated over the course of the twentieth century, theservice sector has expanded faster than manufacturing employment, as rising incomes create moredemand for local services. In this paper, we examine the impact of manufacturing jobs on serviceemployment, estimating local labour multipliers associated with manufacturing jobs in five emerg-ing economies, including Brazil, China, India, Mexico and South Africa. Doing so, we show thatmultipliers are (i) substantially larger relative to previous estimates for advanced economies, and(ii) that by shifting towards human capital-intensive modes of production—as historically achievedby countries such as Singapore, Taiwan and South Korea—emerging economies can significantlyboost job creation in the nontradable sector.

Our findings have important implications for policy. First, becasue local multipliers provideupper bounds for national multipliers for nontradable sectors (Moretti, 2010), our findings aresuggestive of higher national multipliers for nontradables in emerging economies.16 For countriessuch as China, seeking to shift from export-led growth to domestic consumption, the mangnitudeof the multiplier across industries and skill groups is important to guide policymaking. Second,while standard trade models suggest that countries should specailize in goods and services wherethey hold a comparative advantage, a growing body of work shows that countries specializing intechnologically sophisticated products generally grow faster (Hausmann et al., 2007).17 In linewith the literature showing that countries specializing in skill-intensive production grow faster, wedocument that these countries also create more service jobs.

References

Amirapu, A. and Subramanian, A. (2015). Manufacturing or services? an indian illustration of adevelopment dilemma. Center for Global Development Working Paper, (408).

Baer, W. and Samuelson, L. (1981). Toward a service-oriented growth strategy. World Develop-

ment, 9(6), 499–514.

16The reason why local multipliers provide upper bounds for the nontradable sectoer is because the elasticity oflabor supply is lower nationally and the increased possibility that new jobs in the nontradable sector will crowd outjobs in tradables on the national level.

17Rodrik (2006) further shows that the rise of China is not a simple story of specialization according to comparativeadvantage: China produces more technologically sophisticated tradables than its level of income would predict.

12

Page 13: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Bairoch, P. (1988). Cities and economic development: from the dawn of history to the present.University of Chicago Press.

Bartik, T. J. (1992). Who benefits from state and local economic development policies. Journal of

the American Planning Association, 58(2), 258–259.

Center, M. P. (2015). Integrated public use microdata series, international: Version 6.4.

Faggio, G. and Overman, H. (2014). The effect of public sector employment on local labourmarkets. Journal of urban economics, 79, 91–107.

Gollin, D., Parente, S., and Rogerson, R. (2002). The role of agriculture in development. The

American Economic Review, 92(2), 160–164.

Greenstone, M. and Moretti, E. (2004). Bidding for industrial plants: Does winning a million dollarplant increase welfare. Mimeo.

Greenstone, M., Hornbeck, R., and Moretti, E. (2010). Identifying agglomeration spillovers: Evi-dence from winners and losers of large plant openings. Journal of Political Economy, 118(3).

Hausmann, R., Hwang, J., and Rodrik, D. (2007). What you export matters. Journal of economic

growth, 12(1), 1–25.

Hornbeck, R. and Naidu, S. (2014). When the levee breaks: Black migration and economic devel-opment in the american south. THE AMERICAN ECONOMIC REVIEW, 104(3), 963–990.

Jones, P. (2001). Are educated workers really more productive? Journal of development economics,64(1), 57–79.

Kijima, Y. (2006). Why did wage inequality increase? evidence from urban india 1983–99. Journal

of Development Economics, 81(1), 97–117.

Marchand, J. (2012). Local labor market impacts of energy boom-bust-boom in western canada.Journal of Urban Economics, 71(1), 165–174.

Mian, A. and Sufi, A. (2010). The effects of fiscal stimulus: evidence from the 2009’cash forclunkers’ program. Technical report, National Bureau of Economic Research.

Michaels, G., Rauch, F., and Redding, S. J. (2013). Task specialization in us cities from 1880-2000.Technical report, National Bureau of Economic Research.

Morawetz, D. (1974). Employment implications of industrialisation in developing countries: Asurvey. The Economic Journal, 84(335), 491–542.

13

Page 14: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Moretti, E. (2010). Local multipliers. The American Economic Review, 100(2), 373–377.

Moretti, E. (2011). Local labor markets. Handbook of labor economics, 4, 1237–1313.

Moretti, E. and Thulin, P. (2013). Local multipliers and human capital in the united states andsweden. Industrial and Corporate Change, 22(1), 339–362.

Moretti, E. and Wilson, D. J. (2014). State incentives for innovation, star scientists and jobs:Evidence from biotech. Journal of Urban Economics, 79, 20–38.

Park, S.-H. (1989). Linkages between industry and services and their implications for urban em-ployment generation in developing countries. Journal of Development Economics, 30(2), 359–379.

Psacharopoulos, G. (1994). Returns to investment in education: A global update. World develop-

ment, 22(9), 1325–1343.

Psacharopoulos, G. and Patrinos, H. A. (2004). Returns to investment in education: a furtherupdate. Education economics, 12(2), 111–134.

Rodrik, D. (2006). What’s so special about china’s exports? China & World Economy, 14(5), 1–19.

Rodrik, D. (2015). Premature deindustrialization. Journal of Economic Growth, pages 1–33.

Saks, R. E. (2008). Job creation and housing construction: Constraints on metropolitan area em-ployment growth. Journal of Urban Economics, 64(1), 178–195.

Staiger, D. O. and Stock, J. H. (1997). Instrumental Variables Regression with Weak Instruments.Econometrica, 65(3), 557–586.

Stock, J. H. and Yogo, M. (2005). Testing for weak instruments in linear iv regression. In E. D.Andrews and J. Stock, editors, Identification and Inference for Econometric Models: Essays in

Honor of Thomas Rothenberg., pages 80–108. Cambridge: Cambridge University Press.

Stock, J. H., Wright, J. H., and Yogo, M. (2002). A Survey of Weak Instruments and Weak Identi-fication in Generalized Method of Moments. Journal of Business & Economic Statistics, 20(4).

Van Dijk, J. (2014). Local employment multipliers in us cities. Technical report.

Vijverberg, W. P. and Zeager, L. A. (1994). Comparing earnings profiles in urban areas of anldc: Rural-to-urban migrants vs. native workers. Journal of Development Economics, 45(2),177–199.

14

Page 15: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Wilson, D. J. (2012). Fiscal spending jobs multipliers: Evidence from the 2009 american recoveryand reinvestment act. American Economic Journal: Economic Policy, pages 251–282.

A APPENDIX

[Table 6 about here.]

15

Page 16: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

List of Figures1 Share of Tradable and Non-tradable Sector of Total Employment, 1980-2010. . . . 172 Share of Skilled Jobs in the Tradable Sector, 1980-2010. . . . . . . . . . . . . . . 18

16

Page 17: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

0

0.05

0.1

0.15

0.2

0.25

1980 1990 2000 2010

1.A Share of Tradable Sector

South Africa China India Mexico Brazil

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

1980 1990 2000 2010

1.B Share of Non-tradable Sector

South Africa China India Mexico Brazil

(a) Tradable (b) Non-tradableNotes: These figures show the share of urban employment in the tradable and nontradable sector for the five countriesincluded in the analysis, based on data from the IPUMS-I samples.

Figure 1: Share of Tradable and Non-tradable Sector of Total Employment, 1980-2010.

17

Page 18: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

0

0.1

0.2

0.3

0.4

0.5

1980 1990 2000 2010

2.A Share of Skilled Tradable Sector (Secondary)

South Africa China India Mexico Brazil

0

0.02

0.04

0.06

0.08

0.1

0.12

1980 1990 2000 2010

2.B Share of Skilled tradable Secotr (College)

South Africa China India Mexico Brazil

(a) Secondary (b) CollegeNotes: These figures show the share of skilled urban employment in the tradable and nontradable sector for the fivecountries included in the analysis, based on data from the IPUMS-I samples.

Figure 2: Share of Skilled Jobs in the Tradable Sector, 1980-2010.

18

Page 19: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

List of Tables1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 Local Multipliers in Emerging Economies, 1980-2010: OLS estimates. . . . . . . . 213 Local Multipliers in Emerging Economies, 1980-2010: IV estimates. . . . . . . . . 224 Local Multipliers in Emerging Economies, 1980-2010: Non-Durable Tradables on

Durable Tradables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 Local Multipliers in Emerging Economies, 1980-2010: Durable Tradables on Non-

Durable Tradables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246 Local Multipliers in Emerging Economies: IV estimates with different urban pop-

ulation thresholds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

19

Page 20: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Country (available years): Employment in: Mean Std. Deviation Minimum MaximumSouth Africa Tradable 47,394.8 77,957.47 342.93 395,940.42001-2007 Nontradable 212,134.8 330,966 5284.09 1,503,414

Tradable skilled (secondary) 19,401.85 33,679.73 114 169,886.8Tradable unskilled (secondary) 27,742.91 44,000.85 228.39 222,557.3Tradable skilled (college) 2,527.42 5,055.55 0 23,712.95Tradable unskilled (college) 44,617.34 72,734.09 342.93 36,8731.1

China Tradable 314,514.2 366,297 100 2,548,0001982-1990 Nontradable 332,848.6 370,870.5 3,800 3,848,700

Tradable skilled (secondary) 83,707.69 123,064.7 0 1,093,900Tradable unskilled (secondary) 230,806.5 260,726.3 100 1,454,100Tradable skilled (college) 4,668.62 9,737.9 0 87,900Tradable unskilled (college) 309,845.5 359,021.9 100 2,460,100

India Tradable 212,827.4 294,842.4 0 1,801,3621983-1993-2004 Nontradable 555,981.1 660,908.8 4,971.41 4,147,138

Tradable skilled (secondary) 62,653.12 102,472.1 0 689,607.6Tradable unskilled (secondary) 150,109.2 202,582.3 0 1,199,465Tradable skilled (college) 17,802.32 32,864.85 0 218,963Tradable unskilled (college) 194,960 265,392.7 0 1,582,399

Mexico Tradable 10,626.37 29,959.49 84 413,7201990-2000-2010 Nontradable 39,177.01 117,500.4 870 1,972,586

Tradable skilled (secondary) 2,714.62 8,343.45 0 117,970Tradabl unskilled (secondary) 7,845.99 22,043.3 60 295,750Tradable skilled (college) 851.95 2,986.68 0 47,125Tradable unskilled (college) 9,708.66 27,211.26 84 370,350

Brazil Tradable 8,967.91 36,269.64 141.45 1,212,4601980-1991-2000-2010 Nontradable 41,973.91 154,171.5 1,855 4,228,272

Tradable skilled (secondary) 2,618.21 11,310.23 0 299,756.4Tradable unskilled (secondary) 6,305.08 26,217.21 108.48 1,000,260Tradable skilled (college) 490.5 3,192.36 0 93,017.99Tradable unskilled (college) 8,432.79 33,299.12 141.45 1,148,800

Notes: This table reports descriptive statistics for the five countries included in the empirical anal-ysis based on data from IPUMS-I.

Table 1: Descriptive Statistics

20

Page 21: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Local Multipliers (OLS)

Country (observations)All Secondary College

(1) (2) (3) (4)South Africa (24) 1.423*** 1.195*** 3.214*** 15.011***

(0.068) (0.197) (0.367) (2.133)China (129) 0.502*** 0.494*** 1.581*** 27.274***

(0.136) (0.142) (0.494) (10.327)India (146) 1.575*** 1.331*** 3.471*** 6.399***

(0.306) (0.175) (0.771) (2.029)Mexico (944) 0.174 0.842*** 6.655** 14.272**

(0.491) (0.307) (2.791) (5.978)Brazil (2730) -1.622*** -0.863*** 4.123*** 8.639***

(0.165) (0.224) (0.795) (1.328)Controls? N Y Y YTime dummies? Y Y Y Y

Notes: This table presents OLS estimates of equation (1). Each cell corresponds to an individual regression. Controlvariables include: log of total employment, the unemployment rate, the share of female employment, and the shareof employment in age groups 15-24, 25-39, and 40-54 respectively. Statistical significance based on standard errorsclustered at the city-level is denoted by: *** p<0.01, ** p<0.05, * p<0.10.

Table 2: Local Multipliers in Emerging Economies, 1980-2010: OLS estimates.

21

Page 22: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Local Multipliers (2SLS)

Country (observations)All Secondary College(1) (2) (3)

South Africa (24) 1.105*** 2.898*** 15.890***(0.146) (0.288) (1.618)

China (129) 5.122** 18.621* 108.603(2.191) (10.780) (67.958)

India (146) 2.184*** 8.715*** 21.438***(0.430) (1.692) (3.591)

Mexico (944) 6.307*** 29.952*** 91.924(0.755) (11.023) (105.380)

Brazil (2730) 0.496*** 12.474*** 13.125***(0.139) (1.470) (1.790)

Controls? Y Y YTime dummies? Y Y Y

Notes: This table presents 2SLS estimates of equation (1). Each cell corresponds to an individual regression. Controlvariables include: log of total employment, the unemployment rate, the share of female employment, and the shareof employment in age groups 15-24, 25-39, and 40-54 respectively. Statistical significance based on standard errorsclustered at the city-level is denoted by: *** p<0.01, ** p<0.05, * p<0.10.

Table 3: Local Multipliers in Emerging Economies, 1980-2010: IV estimates.

22

Page 23: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Local Multipliers (OLS/2SLS)A. OLS B. IV

Country (observations)All Skilled All Skilled(1) (2) (3) (4)

South Africa (24) 1.147*** 3.518*** 0.704*** 0.824(0.277) (0.795) (0.217) (1.831)

China (129) 0.401** 0.658* 14.230 -17.691(0.166) (0.367) (49.870) (21.793)

India (146) 0.375*** 0.575* 0.258* 1.668*(0.059) (0.311) (0.150) (0.975)

Mexico (944) 0.885*** 0.626** 0.217 -5.205***(0.128) (0.285) (0.142) (0.858)

Brazil (2730) 1.144*** -2.676** 2.080*** -17.956***(0.116) (1.271) (0.311) (2.972)

Controls? Y Y Y YTime dummies? Y Y Y Y

Notes: This table presents OLS and 2SLS estimates of equation (1). Each cell corresponds to an individual regression.Control variables include: log of total employment, the unemployment rate, the share of female employment, and theshare of employment in age groups 15-24, 25-39, and 40-54 respectively. Statistical significance based on standarderrors clustered at the city-level is denoted by: *** p<0.01, ** p<0.05, * p<0.10.

Table 4: Local Multipliers in Emerging Economies, 1980-2010: Non-Durable Tradables onDurable Tradables.

23

Page 24: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Local Multipliers (OLS/2SLS)A.OLS B. IV

Country (observations)All Skilled All Skilled(1) (2) (3) (4)

South Africa (24) 0.481** 1.050** 0.413*** 0.857***(0.185) (0.437) (0.132) (0.287)

China (129) 0.552*** 1.279*** 0.108 -0.139(0.122) (0.435) (0.269) (0.428)

India (146) 1.066*** 1.697*** 2.742*** 7.417(0.221) (0.515) (0.824) (6.141)

Mexico (944) 0.323** 1.065** 1.071** 0.045(0.134) (0.463) (0.438) (1.316)

Brazil (2370) 0.380*** -0.101 0.286*** -0.158(0.071) (0.107) (0.047) (0.164)

Control Variables Y Y Y YTime dummies Y Y Y Y

Notes: This table presents OLS and 2SLS estimates of equation (1). Each cell corresponds to an individual regression.Control variables include: log of total employment, the unemployment rate, the share of female employment, and theshare of employment in age groups 15-24, 25-39, and 40-54 respectively. Statistical significance based on standarderrors clustered at the city-level is denoted by: *** p<0.01, ** p<0.05, * p<0.10.

Table 5: Local Multipliers in Emerging Economies, 1980-2010: Durable Tradables on Non-Durable Tradables.

24

Page 25: Cities, Industrialization, and Job Creation: Evidence from … · 2019-07-29 · Cities, Industrialization, and Job Creation: Evidence from Emerging Economies Thor Berger Chinchih

Tradable (all) Skilled (Secondary) Skilled (College)

100,000+ 500,000+ 1 Million+ 100,000+ 500,000+ 1 Million+ 100,000+ 500,000+ 1 Million+

(1) (2) (3) (4) (5) (6) (7) (8) (9)

South Africa 1.071*** 0.904*** 0.783*** 2.824*** 2.439*** 2.844*** 15.683*** 12.954*** 7.362***

(0.142) (0.174) (0.041) (0.283) (0.400) (0.286) (1.854) (1.790) (0.719)

observations 20 12 6 23 19 6 20 12 6

China 5.122** 5.052** 5.015** 18.621* 18.726* 18.465* 108.603 109.101 109.220

(2.191) (2.177) (2.124) (10.780) (10.976) (10.646) (67.958) (68.103) (68.091)

observations 129 120 115 129 120 115 129 120 115

India 2.182*** 2.192*** 2.138*** 8.710*** 8.842*** 9.261*** 21.455*** 21.801*** 21.952***

(0.429) (0.439) (0.451) (1.695) (1.836) (2.294) (3.615) (3.956) (4.694)

observations 134 117 97 134 117 97 134 117 97

Mexico 5.520*** 4.782*** 7.785** 33.147* 16.726 9.315 65.260 284.159 88.597

(0.730) (0.835) (3.356) (17.913) (21.387) (12.377) (217.070) (1,763.065) (275.654)

observations 248 64 20 248 64 20 248 64 20

Brazil 0.561*** 0.715*** 0.768*** 11.319*** 8.180*** 6.167*** 11.377*** 8.866*** 7.688***

(0.145) (0.157) (0.150) (2.144) (1.722) (0.561) (1.788) (0.928) (0.405)

observations 716 118 46 716 118 46 716 118 46

Notes: This table presents 2SLS estimates of equation (1) where the samples are restricted to cities with at least100,000, 500,000, and 1 million inhabitants respectively. Each cell corresponds to an individual regression. Controlvariables include: log of total employment, the unemployment rate, the share of female employment, and the shareof employment in age groups 15-24, 25-39, and 40-54 respectively. Statistical significance based on standard errorsclustered at the city-level is denoted by: *** p<0.01, ** p<0.05, * p<0.10.

Table 6: Local Multipliers in Emerging Economies: IV estimates with different urban populationthresholds.

25