1 COMPETITIVE CITIES FOR JOBS AND GROWTH 101718 COMPANION PAPER 1 CITY ANALYTICS Kenan Fikri, T. Juni Zhu December 2015 2 TABLE OF CONTENTS Background and Acknowledgments 5 Executive Summary 6 1. Introduction 8 2. City Profile and Performance 8 What data on cities are available? 8 Exploring the Oxford Economics data 8 Profile of the average city in the data set by region 9 Global variation in the data set 10 Comparing city and national economic performance 12 Productivity differentials across cities and within countries 14 Private sector job creation 15 Industrial sector job creation 16 Tradable sector job creation 17 Foreign direct investment 18 3. Development Pathways: How Do City Economies Evolve over Time? 20 Income pathway 20 City GDP per capita growth by region 21 City GDP per capita growth by region, controlling for country effects 23 City GDP per capita growth by income level 24 City GDP per capita growth by income level, controlling for country effects 25 City GDP per capita growth by industrial profile 26 Income traps 26 Industry pathways 28 Size pathways 31 4. Predictors of City Competitiveness 33 Working definition of city competitiveness 33 Stocktaking of city indexes 33 How well do city indexes predict competitiveness outcomes? 34 City competitiveness correlates: Factor analysis 35 Factor analysis: Global results on the determinants of city competitiveness 35 Predictors of city competitiveness by region and city type 37 How does the “mayor’s wedge” relate to city competitiveness? 39 Findings from the mayor’s wedge analysis: Scope and autonomy 41 Findings from the mayor’s wedge analysis: Capacity 41 Variation in the impact of the mayor’s wedge by region and city type 42 How does inequality relate to city competitiveness? 44 5. Attempting to Arrive at a List of the World’s Most Competitive Cities 45 6. Conclusion 46 References 48 3 4 BACKGROUND AND ACKNOWLEDGEMENTS I nterest in studying city competitiveness has skyrocketed income, sector, region, and so on. And we have buttressed in the past few years, although the topic itself is far from these findings with econometric deep dives and case stud- new. Mayors and city leaders have long worried about ies in selected countries and cities. We are able to inform the obstacles to job creation, competitiveness, and economic the ongoing debates on what really matters for economic growth that plague their cities. outcomes in cities with analysis of overarching trends and associations, supplemented with rigorous analyses to identify This paper is part of a broader research initiative, the Com- causal relationships. We also try to “stand on the shoulders of petitive Cities Knowledge Base, which is managed jointly by giants” where possible: that is, we use and reference exist- the Trade and Competitiveness Global Practice and the Social, ing resources (research, analysis, toolkits, and experts). The Urban, Rural, and Resilience Global Practice of the World summary findings of the overall research are presented in Bank Group. Its objective is to create a knowledge base on the framework report, Competitive Cities for Jobs and Growth competitive cities, to improve the understanding of job cre- (World Bank, 2015). ation at the city level and as a foundation for a community of practice on this topic for World Bank staff, academia, donor The objective of this paper is to present key findings from partners, and practitioners. the quantitative analysis of the drivers of competitiveness in cites around the world. Our attempt in this initiative has been to focus our energies on bringing to our clients a robust body of knowledge that This note was prepared by Kenan Fikri and T. Juni Zhu, with will address their questions on benchmarking their perfor- contributions from Anca Rusu and guidance and assistance mance, on understanding what has worked elsewhere and from Austin Kilroy and Megha Mukim. The joint task team what has not, and on looking at ways to organize for delivery leaders of the Competitive Cities Knowledge Base project in different contexts. are Austin Kilroy and Megha Mukim. Stefano Negri, Sameh Wahba, Ceci Sager, and Somik Lall have provided overall Our approach has focused on using different methodologies guidance on the project as senior advisers. to tackle these questions. These methodologies are based on best practices, data availability, replicability, and simplici- The team would like to acknowledge gratefully the European ty. In many cases, we have leveraged new and existing data Commission; the African, Caribbean, and Pacific Group of sources to shed new light on some unanswered questions; States Secretariat; and the governments of Austria, Norway, in others, we have conducted primary research because and Switzerland for financing this study through the Com- available data were inadequate. We looked at global and petitive Industries and Innovation Program. regional trends, comparing different typologies of cities by 5 EXECUTIVE SUMMARY C Key findings from the global quantitative analytics work ities today are home to 54 percent of humanity (UN- DESA 2015). They cover only a fraction of the world’s stream are as follows: landmass but concentrate 80 percent of global gross do- mestic product (GDP) (World Bank 2013).1 Since the Industri- • The story of the world’s cities from 2000 to 2012 was one al Revolution, country after country has witnessed urbaniza- of rising prosperity. Powered by China, GDP per capita tion lift entire generations out of poverty and into the middle across cities in the data set grew by 4.6 percent on aver- class. Cities combine their societies’ resources into something age each year. Of the 750 cities under study, 36 percent more than the sum of their parts. For countries, cities are the achieved annual average growth rates above 5 percent. engines of productivity and the workhorses of development. GDP per capita fell in 46 cities under study—a small For individuals, cities mean opportunity. fraction of the total that includes not only the reces- sion-hit cities of the developed world but also struggling Yet what makes a city thrive and its residents prosper—in cities in Sub-Saharan Africa and beyond, thus reminding short, what makes a city competitive—remains a frustrat- us that development does not always proceed as linear ingly difficult question to answer. To some extent, every city advancement. has its own secret sauce. But some sauces do seem to be more potent than others, a fact that suggests certain ingredients • The world’s cities remain extremely heterogeneous, and and combinations may matter more than others in determin- the results presented here argue emphatically against a ing economic outcomes. This paper aims to track down some one-size-fits-all approach to city competitiveness. The elusive common threads by assessing the state of cities in the most populous city in the data set had a population 144 world today and identifying factors that appear to be correlat- times that of the smallest, and the richest city enjoyed ed with better economic outcomes. We explore these issues a GDP per capita level 384 times that of the poorest. Al- globally, across all cities, as well as within regions and across though China’s cities achieved astonishing growth rates, different city typologies. The paper is designed to summarize income levels and living standards fell in parts of Europe the project’s findings and demonstrate how the database and and Central Asia and of Sub-Saharan Africa.2 Even where analytic tools can be used in other contexts—for example, in growth was robust, productivity, incomes, and employ- regional deep dives. ment were not guaranteed to follow suit. The analysis presented here begins by benchmarking the • Cities generate a disproportionate share of new private economic performance of 750 cities across the globe. It as- sector jobs. From 2006 to 2012, 750 of the world’s sesses not only how quickly those cities are growing but also largest cities analyzed here created 87.7 million private how their economies are changing, how their performance sector jobs, accounting for 58 percent of all new private compares with that of peers at home and abroad, and where sector jobs in their 129 countries over the period despite these cities fit in their national urban systems. We draw from collectively being home to only one-quarter of total these findings common development pathways and then plot private sector employment. Beijing and Chongqing (in individual cities along them. The explanatory portion of the China) and Jakarta (in Indonesia) created the most pri- analysis starts by taking stock of the existing knowledge base vate sector jobs in sheer numeric terms, with more than on city competitiveness with a review of secondary sourc- 2 million new jobs each. es. It proceeds to explore the factors associated with better competitiveness outcomes across multiple posited buckets of • The competitiveness opportunity is huge. If each city had determinants through a simple correlation analysis. Then we elevated the local rate of job creation to that achieved by use regression analysis to explore how facets of local govern- the average city in the top quarter of performers in its ment autonomy relate to city competitiveness. We find no region, 18.9 million extra jobs would have been created holy grail for city competitiveness, but we do find many hints. in 2012 alone. 6 • The manufacturing sector is slowing as an engine of services sector grows to dominate the economy. These growth, particularly for low-income cities. The rate of creative and financial services hubs thrive on human job growth in the industrial sector of the economy has capital and innovation. slowed significantly for low-income countries over the past 12 years, prompting the question whether the • Cities rarely change their position in a national urban services sector will prove to be as powerful an engine of hierarchy except in large countries. National urban development. hierarchies typically exhibit remarkable stability, but when changes do occur they are either dramatic (for • Strong tradable sectors characterize competitive cities. example, New Orleans or Detroit in the United States) In the 10 percent of cities in which GDP per capita grew or in fast-growing countries with large urban systems fastest from 2005 to 2012, tradable sector employment (for example, China, India, or Nigeria). In five countries, growth outstripped non-traded sector employment the primary city in terms of population was displaced by growth by 2.5 percent on average annually. By contrast, 2012 by the city that ranked highest in 2000 in terms of in less competitive cities, tradable and non-tradable GDP. industries grew, both at a slower rate. • The factors associated with positive competitiveness • Foreign direct investment (FDI) remains highly con- outcomes in cities vary by income level. Institutions and centrated in a relatively small number of elite cities. regulations matter at all levels of income and economic However, in terms of FDI inflows per dollar of GDP from structures, whereas physical infrastructure appears to 2003 to 2012, two-thirds of the top FDI destination boost growth at low income levels; social infrastructure cities could be found in Sub-Saharan Africa, South Asia, supports productivity at middle income levels; and inno- and East Asia and Pacific (excluding China). Nonetheless, vation, human capital, and financial infrastructure all the economic development potential of FDI should not contribute to growth, productivity, and living standards be overstated. In every city, local firms still create the in high-income cities. majority of jobs. • Competitive cities are good at mopping up inequality. • The dominant economic sector in a city changes only Even without significant increases in administrative over very long time horizons. When it does change, cities powers, competitive cities—especially cities that attract typically—but not always—follow a well-established a large influx of migrants—are good at reducing inequal- pathway. Four of five cities saw no change in the largest ity. Although global city-level data confirm that in-mi- sector of their economy from 2000 to 2012. Of those gration contributes to intracity inequality in the short that did change, almost half transitioned from industry term, we observe that inequality tends to decline over to high-end services. Several South Asian cities tran- time as cities develop and new migrants are absorbed sitioned from high-end services to consumer services, into the labor force. thus challenging the assumption of a linear development pathway. • Maximizing the quality and pace of economic develop- ment may call for a nuanced approach to devolution. • As incomes rise and cities progress along the develop- Regression analysis found that expanding the scope of a ment pathway, their functions evolve. Up to approxi- mayor’s administrative remit is associated with improved mately US$2,500 GDP per capita, cities serve primarily competitiveness in cities, but the same is not true for as market towns and central points for the exchange financial autonomy. In-depth research carried out in of basic services. At GDP per capita levels of about China suggests that local government capacity is also a US$2,500 to US$12,000, industry dominates the econ- determining factor. omy, and cities take on a new role as production centers and agglomerations of people, capital, and suppliers. As cities progress to higher income levels, the tertiary 7 1. Introduction 2. City Profile and Performance This technical note concluding the global quantitative analyt- What data on cities are available? ics portion of the Competitive Cities Knowledge Base project at the World Bank Group begins by reviewing the available Numerous organizations have attempted to rank cities data on cities and analyzing the recent economic perfor- across the world on one combination of metrics or another, mance of cities across regions and select typologies. It then but city-level data sets with detailed raw data on city eco- presents the findings from correlation and regression analy- nomic structure and performance over time and with global ses studying the predictors of city competitiveness. Finally, coverage are few and far between. The World Bank Group this paper attempts to identify development pathways and purchased a comprehensive data set for this project: the plot cities along them to help inform advisory or technical Global Cities Historic Database from Oxford Economics (OE). work across World Bank Group stakeholder communities in Euromonitor International maintains an alternative data- the future. base called Passport: Cities that offers a comparable range of indicators for 126 large cities but a much more limited range The technical, descriptive, and analytic findings discussed (covering population and household characteristics) for more in this paper are intended to uncover global trends, given than 1,000 smaller cities. The Euromonitor data set extends available data, while at the same time illustrating a city-based back only to 2005 but does offer forecasts to 2020 on certain approach to development analysis. The work presented here indicators. can be tailored or deepened in whole or in part to explore city performance across regions or topic areas. The OE database used in this paper covers 750 cities (defined as metropolitan areas) across 140 different countries. The data set is not intended to be comprehensive; the cities were selected from the United Nations list of urban agglomer- ations with at least 750,000 inhabitants and then supple- mented with other strategic cities such as country capitals.3 The data set contains 12 years of historical data, covering the period from 2000 to 2012, and includes 90 different variables covering demographics, output and employment (each by sec- tor), household income, consumer spending, and retail sales, among others. This analysis uses only the small number of indicators that deal directly with economic outcomes. Findings presented here are derived from the OE database unless otherwise stated. Additional sources include data on public finances (revenues and expenditures) that were obtained from the International Monetary Fund and patent- ing data from the Global Urban Competitiveness Project. See tables 4.1 and 4.2 for a list of the variables and their sources that were used in the correlation and regression analysis. Exploring the Oxford Economics data Even within OE’s sample of the world’s 750 largest cities, the basic characteristics of cities vary hugely across regions, and before we dive into the analysis, it will be instructive to take stock of the heterogeneity. Table 2.1 provides a profile of the average city in each World Bank Group region contained in the data set.4 The n reports the number of cities from each region. 8 Table 2.1 Profile of the average city in each World Bank Group region Source: Oxford Economics Dataset Note: GDP = gross domestic product; GVA = gross value added. Profile of the average city in the data set by percent of employment and nearly 50 percent of gross value region added (GVA)—far above the average in any other region. Consumer services and high-end services sectors were under- In 2012, the average Organisation for Economic Co-operation sized. and Development (OECD) city in the data set was a high-in- come city of 2.8 million inhabitants that had extremely high From Europe and Central Asia, the average city in the data levels of labor productivity by global standards that was set was upper-middle income with 1.4 million people. Many growing slightly faster than its national economy.5 From fig- cities dominated their national economies, with the average ure 2.1, we see that high-end services dominated the average city generating over one-tenth of country GDP. In terms OECD city economy, accounting for almost 40 percent of of population, European and Central Asian cities were the value added. slowest growing in the world. In terms of employment and GDP, they grew slightly faster than their national economies The average East Asian and Pacific city in the data set in 2012 in 2012. was an upper-middle-income city with 5.5 million people— more than in any other region—and labor productivity of The average city of the data set in the Middle East and US$33,000 in value added per worker per year, second behind North Africa was upper-middle income with 2.0 million OECD cities. Job growth rates in East Asian and Pacific cities people. Such a city had relatively high labor productivity, at exceeded their national averages by 1.5 percent and output US$30,500 in value added per worker per year, and a relative- growth rates by 2.1 percent. Industry and mining dominated ly high population growth rate. The average city in the Middle the economy of the average city and accounted for over 35 East and North Africa played a significant role in its national Figure 2.1 Industrial profile of the average city in each region, 2012 Share of gross value added (%) Share of city employment (%) Share of GVA (%) Share of city employment 0 10 20 30 40 50 Sector 60 70 share 80 90of 100 average city employment 0 10 20 30 40 50 60 70 80 90 100 East Asia and Pacific East Asia and Pacific East Asia and Pacific Middle East and North Africa Middle East and North Africa Middle East and North Africa Europe and Central Asia Europe and Central Asia Europe and Central Asia Latin America and the Caribbean Latin America and the Caribbean Latin America and the Caribbean South Asia South Asia South Asia Sub-Saharan Africa Sub-Saharan Africa Sub-Saharan Africa OECD OECD OECD Industry Agriculture High-end services Industry Agriculture High-end services Industry Agriculture High-end services Consumer services Public sector Source: Oxford Economics Dataset Note: OECD = Organisation for Economic Co-operation and Development. 9 economy, accounting for 12 percent of total GDP. It also had Global variation in the data set a large public sector—smaller only than OECD cities—and a comparatively small consumer sector. The cities in this global sample vary greatly, and their diver- sity serves as a useful reminder that this paper analyzes 750 The average Latin American and Caribbean city in the data heterogeneous units. Cities differ across many dimensions: set was upper-middle income with 2.3 million inhabitants size, economic vocation, geographic location, natural en- and a relatively young demographic profile. Consumer ser- dowments, income level, history, planning model, political vices employed the largest share of workers, but industry and system, and so on. Furthermore, each city strives to solve high-end services accounted for the largest shares of value dramatically different issues with different starting points. added. A city in Europe and Central Asia may be trying to manage deindustrialization, whereas in South Asia or Sub-Saharan In South Asia, the average city was lower-middle income with Africa, a city may be struggling to integrate new migrants 2.7 million inhabitants. In a region with relatively well-de- into the job market, social fabric, or built environment. In veloped urban systems, the typical city accounted for only OECD countries, a city may be preoccupied with attracting 1.2 percent of its national GDP. In productivity, workers still talent through amenities or with rekindling fading entrepre- lagged far behind their peers in all other regions, including neurial zeal. Sub-Saharan Africa. Labor productivity in the average South Asian city stood at US$6,700 in value added per worker per Figures 2.2 and 2.3 underscore this point. The largest city in year. the data set, Tokyo (Japan), has 144 times the population of the smallest, Gaborone (Botswana). Even more starkly, the Finally, the data set’s typical city in Sub-Saharan Africa was richest city in the data set, Basel (Switzerland), enjoys a GDP lower-middle income with. 2.0 million people. The region per capita that is an astonishing 384 times higher than that itself experienced faster population growth than any other of the poorest city, Kinshasa (Democratic Republic of Congo). region. The average city in Sub-Saharan Africa was adding jobs and increasing output at a significantly faster rate than Nor are the cities different only across static measures. The its country, and the typical city’s impact on national eco- four panels of figure 2.3 show the maximum, minimum, nomic performance mattered more in Sub-Saharan Africa and mean city performance across four economic indicators than elsewhere because the typical city accounted for over 15 for the period from 2000 to 2012. While in the United Arab percent of its country’s GDP. Emirates, Dubai’s population grew by 11.3 percent per year, in Latvia, Riga’s population fell by 1.5 percent. Putting the competitiveness imperative in stark relief, one notes that as Figure 2.2 Spectrum of city starting points within the average household disposable income grew by 13.7 percent per year in Guigang (China), it fell by 6.0 percent per year in Oxford Economics data set Yamoussoukro (Cameroon). These disparities are not entirely driven by outliers. The People (millions) GDP per capita (US$, thousands) difference between the average city in the top and bottom 10 40 120 percent globally on these metrics ranged from 5.2 percent for 36.7 108.2 population growth to 13.7 percent for GDP growth. Popula- 35 100 tion, employment, and living standards all shrank from 2000 30 to 2012 in the bottom 10th of the world’s cities analyzed here. 80 25 20 60 15 40 10 20 mean 17.1 5 mean 3.2 0.2 0.3 0 0 Minimun Maximun Minimun Maximun Source: Oxford Economics Dataset 10 Figure 2.3 Measuring the disparities in city performance observed in the data set, 2000-2012 PANEL A PANEL B PANEL C PANEL D Average Annual Growth, Average Annual Growth, Average Annual Growth, Average Annual Growth, Population Jobs GDP Household Disp. Income Percent Percent Percent Percent 25 25 25 25 20 20 20 20 15 15 15 15 23% 10 10 10 gap 10 12.8% 17% mean 5 gap 5 gap 5 mean 5 20% mean mean gap 0 0 0 0 -5 -5 -5 -5 -10 -10 -10 -10 Minimun Maximun Minimun Maximun Minimun Maximun Minimun Maximun Source: Oxford Economics Dataset 11 Comparing city and national economic formed their national economies and peers. These cities are performance of particular interest because of their particular capacity to achieve high growth rates no matter the characteristics of At the country level, identifying outlier cities is the first step their national economies. in determining what differentiates the most competitive cities within their peer group. The city profiles from table Figure 2.4 compares the average difference between city 2.1 show that across regions the average city outperformed and country employment growth each year over the 12-year its country in terms of job growth from 2011 to 2012, and period. Several pockets of over- and underperformance stand except for the average city in Latin America and the Carib- out. Cities performed much better than their national econo- bean and the Middle East and North Africa, this metric also mies in parts of West Africa centered on Nigeria, in East Af- held for GDP growth. Certain individual cities far outper- rica, in coastal and inland China, in parts of Bangladesh and India, and in portions of Southeast Asia. Cities in Europe; Figure 2.4 Annual city employment growth rates relative to national economy, 2000–12 average Source: Oxford Economics Dataset Figure 2.5 Annual city GDP growth rates relative to national economy, 2000–12 average Source: Oxford Economics Dataset 12 the Russian Federation; Table 2.2 lists the top 10 outperformers from each region Cities in Latin America much of the United over a shorter period—from 2006 to 2012. The top 10 were and the Caribbean and States; and portions of determined by summing each city’s average annual employ- Australia, Latin Amer- ment and GDP growth rates above the national average over the Middle East and ica, the Arab Republic the same time period. The shorter period was chosen (a) to North Africa were less of Egypt, Japan, South reduce the number of economic cycles that the measurement likely to outperform their Africa, and northeast cut across and (b) because data before 2006 are unavailable national economies than China underperformed for some cities. Chinese cities have been excluded to let other peers from other regions. their national econo- outperformers in East Asia and Pacific rise to the top. The mies—in some cases, table shows that the vast majority of cities in South Asia and significantly so. Sub-Saharan Africa outperform their national economies. This outcome stands in stark contrast to the situation in The output map (figure 2.5) tells a largely similar story, with the OECD, where fewer than half of cities outperform their a few exceptions: joining the underperformers were a number national economies in terms of GDP. Outperformers are of cities in the Islamic Republic of Iran, the Philippines, and relatively rarer in Latin America and the Caribbean and the Turkey. This time cities in northeast China far outperformed Middle East and North Africa. the country, and European cities did better in terms of GDP as well. Table 2.2 Top 10 outperformers by region, ranked by combined average annual outperformance on output and jobs (2006–12) East Asia & the Europe & Latin America & Middle East & Sub-Saharan Rank Pacific Central Asia the Caribbean North Africa OECD South Asia Africa Nay Pyi Taw, Makhachkala, Cancún, Mexico Sharjah, United Austin, United Pondicherry, Onitsha, Nigeria 1 Myanmar Russian Feder- Arab Emirates States India ation Kota Kinabalu, Sofia, Bulgaria Santa Cruz, Algiers, Algeria Portland Oregon, Patna, India Enugu, Nigeria 2 Malaysia Bolivia United States Vientiane, Lao Tbilisi, Georgia Saltillo, Mexico Muscat, Oman Aberdeen, Unit- Tiruppur, India Abuja, Nigeria 3 PDR ed Kingdom Mandalay, Krasnodar, Rus- Cochabamba, Oran, Algeria San Jose, Cali- Surat, India Benin City, 4 Myanmar sian Federation Bolivia fornia, United Nigeria States Yangon, Myan- Kazan, Russian Teresina, Brazil Marrakesh, Houston, Texas, Ghaziabad, India Abeokuta, 5 mar Federation Morocco United States Nigeria Ujung Pandang, Bucharest, Campo Grande, Riyadh, Saudi Perth, Australia Raipur, India Ilorin, Nigeria 6 Indonesia Romania Brazil Arabia Hanoi, Vietnam Chisinau, Mol- San Luis Potosi, Constantine, Edmonton, Hyderabad, India Ogbomosho, 7 dova Mexico Algeria Canada Nigeria Pekan Baru, Yerevan, Ar- Recife, Brazil Abu Dhabi, Salt Lake City, Bangalore, India Kano, Nigeria 8 Indonesia menia United Arab United States Emirates Ho Chi Minh Penza, Russian Curitiba, Brazil Meknes, Mo- Calgary, Canada Vijayawada, Ibadan, Nigeria 9 City, Vietnam Federation rocco India Ulaanbaatar, St. Petersburg, Chihuahua, Mosul, Iraq San Antonio, Coimbatore, Aba, Nigeria 10 Mongolia Russian Feder- Mexico Texas, United India ation States Number of aver- age annual GDP 40 out of 56 51 out of 71 53 out of 94 22 out of 46 75 out of 176 78 out of 88 63 out of 69 outperformers Number of aver- age annual jobs 40 out of 56 50 out of 71 69 out of 94 35 out of 46 93 out of 176 71 out of 88 59 out of 69 outperformers 13 Source: Oxford Economics Dataset Productivity differentials across cities and average in only one city within countries in the database—large, In a country with a inland Chongqing. Out- relatively mature urban Worldwide, 70 percent of cities were more productive than put per worker is nearly system, such as the United their countries as a whole in 2012 (measured as GVA per six times the national worker). Considerable variation can be found within coun- average in Jilin (Jilin) States, city productivity tries, however. In a country with a mature urban system, and Weihai (Shandong) runs below the national such as the United States, city productivity runs below the and approaches seven level in nearly as many national average in nearly as many cases (23 in this data set) times the national cases as it runs above. as it runs above (27). Output per worker in San Jose (Califor- average in Tangshan nia), the most productive U.S. city, is more than twice that of (Hebei) and Maanshan the least productive, Buffalo (New York), at over US$210,000 (Anhui), where it rises above US$70,000 per year. per worker compared with US$87,500. In rapidly developing China, by contrast, output per worker is below the national Figure 2.6: City versus country productivity levels, 2012 a. All cities vs. all countries United States City GVA/Worker (US$, thousands) China Russian Federation Country GVA/Worker (US$, thousands) b. Cities with less than US$50,000 GVA per worker vs. countries with less than US$25,000 country GVA per worker China City GVA/Worker (US$, thousands) Russian Federation Indonesia Brazil Mexico Colombia India Country GVA/Worker (US$, thousands) Source: Oxford Economics Dataset 14 The OE data show that although cities are typically the Private sector job creation productive engines of their national economies, many exceptions exist. Large (and typically rural-based) natural The private sector in the OE cities created 87.7 million jobs resource sectors change the story in some countries: cities over the six years from 2006 to 2012 (range chosen because such as Sharjah (United Arab Emirates) and Jeddah, Mecca, 2006 is the first year for which all 750 cities report data). and Medina (all in Saudi Arabia) register productivity levels These cities accounted for 58 percent of all new private sector only slightly above half the national value. City productivity jobs created in their 140 countries taken together, despite can lag in countries such as Indonesia and Nigeria, where containing only one-quarter of total private employment. economic development is progressing rapidly nationally but unevenly across subnational regions. And, of course, not all In absolute terms, Beijing and Chongqing (in China) and Ja- cities power economic development at all times; a city such karta (in Indonesia) created the most private sector jobs over as Buffalo, which the period, Jakarta and Beijing each with nearly 2.9 million prospered in one era and Chongqing with 2.3 million. China was home to 10 of the Individual city productivity of industrial develop- 12 cities that created more than 1 million private sector jobs can lag in countries such ment, may experience each. Outside East Asia and Pacific, Lagos, Nigeria, saw the as Indonesia and Nigeria, relative decline in the greatest private sector job creation, with 1.5 million new pri- where economic develop- next. Multiple such vate jobs in six years, followed by São Paulo, Brazil (950,000 ment is progressing rapidly forces appear to be jobs); Bogotá, Colombia (816,000 jobs); Lima, Peru (807,000); at work in Russia, Dhaka, Bangladesh (766,000); and Dar es Salaam, Tanzania nationally but unevenly (754,000). In Europe and Central Asia, Moscow, Russia, saw where only one-fifth of across subnational regions. cities, led by Moscow the greatest absolute increase in private sector jobs (617,000 and Tyumen, produce jobs), whereas Tokyo, Japan, led OECD countries (525,000 greater output per jobs). All together and in absolute terms, the 7 percent of cit- worker than the national economy. Russia is, in fact, home to ies where job creation was most voluminous created as many four of the five cities with the lowest ratio of city to national private sector jobs as the remaining 93 percent combined. For productivity in the world—Penza, Orenburg, Makhachkala, context, these cities were home to just over one-quarter—27 and Barnaul. percent—of total employment in the sample in 2005. Panel b of figure 2.6 takes a closer look at the distribution In percentage terms, China claimed 39 of the 50 cities with of city productivity levels within countries where GVA the fastest average annual rates of private sector job growth, per worker is less than US$25,000 annually. That national which ranged from 17.0 percent in Hefei (Anhui) to 7.5 productivity in China still trails so far behind productivity percent in Baotou (Inner Mongolia). Other cities in the top 50 in the country’s large cities betrays just how much urban- were oil-rich city-states, such as Doha (Qatar) and Abu Dhabi ization has coincided with productivity gains. Most cities in and Sharjah (United Arab Emirates), as well as Myanmar’s India are also more productive than the national economy, newly created capital, Nay Pyi Taw. Six Nigerian cities, includ- but the distribution is far more compressed: productivity in ing Onitsha, Enugu, highest-ranking Delhi and Abuja, joined the is only three-and-one- top 50, as did Dar es Developing countries may That national productivity half times productivity Salaam, Tanzania. be threatened by a process in China still trails so far in lowest-ranking Va- Santa Cruz (Bolivia) of premature deindustri- behind productivity in its ranasi. In comparison, posted the fastest pri- large cities betrays just in Indonesia, Pekan vate sector job growth alization whereby man- Baru is more than five in Latin America and ufacturing’s share of the how much more urbaniza- times as productive as the Caribbean (5.8 per- economy begins to decline tion the country stands to Yogyakarta (and seven cent), Tbilisi (Georgia) at much lower income levels undergo. times as productive as the fastest in Europe than ever before. Samarinda, an outlier). and Central Asia (5.6 Brazil and Mexico ex- percent), Pondicherry hibit a more balanced distribution of cities above and below (India) the fastest in the national average. In Mexico, Monterrey stands out with South Asia (5.0 percent), and Perth (Australia) the fastest in US$39,800 GVA per worker, ahead of the capital, Mexico City OECD countries (3.2 percent). All together and in relative (US$30,800), and more than double the national average of terms, the 16 percent of cities that grew jobs the fastest over US$23,800. Colombia, one of the recent past’s development the period created as many private sector jobs as the bottom success stories, boasts only two cities more productive than 84 percent combined. For context, these cities were home to the national economy as a whole (US$16,300 GVA per work- only 14 percent of total sample employment in 2005. er): Bogotá and Bucaramanga. Mainly as a result of the recession and Euro Area crisis, pri- vate sector employment declined across half of OECD cities over the period. The largest absolute declines took place in 15 U.S. cities hit hard by the housing crisis, such as Los Angeles Industrial sector job creation (California), Phoenix (Arizona), and Riverside (California); in cities of struggling Euro Area economies, such as Athens First, a definitional note: OE groups the mining and ex- (Greece), Dublin (Ireland), Lisbon (Portugal), and Madrid traction, manufacturing, utilities, and construction indus- and Barcelona (Spain); and in industrial cities such as Osaka tries together into its “industrial” sector aggregate, and it (Japan), Birmingham (United Kingdom), and Chicago (United does not break out data for any of these individual subsec- States). In percentage terms, private sector employment fell tors. We refer to the sector as the industrial or manufacturing fastest in struggling Euro Area and in U.S. Sun Belt cities. sector, keeping in mind that the measure itself is broader and that its composition likely varies across cities. When their rates of private sector job growth are compared to the national rate, Chinese cities again come out on top. We take a special look at industry because of manufacturing’s Attesting to the incredible pace and scale of urbanization in historical role as a ladder for economic development. The ex- the country, nationwide the number of private sector jobs port-led growth model that has propelled so many countries grew by only 0.2 percent a year from 2006 to 2012—a rate up the income ladder is at its core a labor-intensive, manufac- that 138 of the 150 Chinese cities in the dataset beat easily. turing-led growth model, in which countries take advantage Tbilisi (Georgia), Kumasi (Ghana), Pekan Baru (Indonesia), of their low initial labor costs (the primary variable cost in Onitsha (Nigeria), and Dar es Salaam (Tanzania) also regis- low-end manufacturing) to attract export-oriented produc- tered private sector job growth rates far above (5.0 percent or tion capital. The heavily urbanized sector absorbs population higher) their countries. from the hinterlands, and this steady flow of labor from the countryside keeps wages low. Eventually this process slows Boosting city competitiveness has the potential to drastical- as urbanization rates taper off and wages begin to rise, but at ly accelerate job creation globally. Cities perform unequally that point companies have typically spent years learning and within regions. For example, cities in the top quartile of building the capacity to move to higher-value-added activi- Sub-Saharan African cities created jobs 4.5 percentage points ties. In this way, industrialization and urbanization unleash a faster than did their peers in the region. In each of the other positive development spiral. regions, the gap between the top quarter and the rest stood at 3.0 percentage points or more in 2012. If all cities in a region Development economist Dani Rodrik (Rodrik, D. (2015). Pre- grew at least as fast as the average city in the top quartile, mature Deindustrialization (No. w20935). National Bureau however, 18.9 million extra jobs—on top of the 13.4 million of Economic Research), however, raises the disturbing specter actually witnessed—would have been created in 2012 alone. that the export-led growth model may have run its course. Figure 2.7 Industrial sector rates of job growth by Figure 2.8 Industrial sector rates of job growth by city income bracket, 2001–12 city specialization, 2001–12 Percent Percent 10 10 8 8 Low-income cities 6 6 Consumer services cities Lower-middle-income cities Industrial cities 4 4 2 2 Upper-middle-income cities 0 0 -2 -2 -4 -4 High-end services cities -6 -6 High-income cities -8 -8 Source: Oxford Economics Dataset Source: Oxford Economics Dataset 16 Several converging trends concern him. First, manufactur- If we compare cities by region, those in the Middle East and ing today is more capital- and skill-intensive than it used North Africa typically post the fastest industrial sector job to be and is growing less labor-intensive even at lower ends growth rates, a finding that is presumably tied to the price of of the value chain as rote automation technologies become oil and the natural resources component of the aggregate sec- less expensive. Second, the rapid growth in international tor. Cities in Sub-Saharan Africa began the study period with logistics has segmented supply chains to an extraordinary high growth rates averaging 5.8 percent from 2001 to 2004 degree, thereby enabling companies to separate discrete tasks before falling toward zero in 2009 and rising back to about across locations rather than move the entire process—from 2.0 percent in 2012. South Asian cities have seen the sector’s which countries could learn—as in previous waves of moving job growth rate decline more or less steadily from over 5.0 manufacture offshore. High-value-added activities, mean- percent early in the decade to 2.0 percent in 2012, while cities while, remain in the developed world. Third, global financial in Europe and Central Asia and especially OECD cities have markets encourage and enable “plug in, plug out” behavior struggled to maintain positive growth rates throughout the whereby companies pick up and move on to the next, cheaper, period. location at the slightest change in cost calculus. In such a world, any country that begins to move up the value chain Tradable sector job creation will have the proverbial rug pulled out from under it as the foreign capital supporting its industrial base departs swiftly. Strong tradable sectors characterize competitive cities. In Compounding the problem is Rodrik’s fourth concern: that the OE data, the tradable sector can broadly be defined as the China’s unprecedented scale crowded other countries off the industrial sector (manufacturing, mining, and construction); development ladder just when they could have gotten ahead the transportation and warehousing, communications, and of these dynamics. information sectors; and the financial and business services sectors combined. In the 10 percent of cities in which GDP As a consequence, goes the hypothesis, developing countries per capita grew fastest from 2005 to 2012, tradable sector may be threatened by a process of premature deindustrializa- employment growth outstripped non-traded sector employ- tion whereby manufacturing’s share of the economy begins ment growth by 2.5 percent on average annually: 6.2 percent to decline at much lower income levels than ever before. growth compared with 3.7 percent growth. By contrast, in Whether traded services will be able to generate the surplus, the remaining 90 percent of cities, tradable and non-tradable productivity gains, and investment capital needed to drive industries grew at effectively the same rate: respectively 2.4 the economic growth as manufacturing historically has done percent and 2.3 percent on average annually. Thus, robust remains to be seen. tradable sector growth characterizes competitive cities and appears to be a prerequisite for rapid growth in the economy The data presented in figure 2.7 lend tentative support to as a whole. the deindustrialization hypothesis. At the beginning of the 2000s, cities in low-income countries registered average Faster growth in the traded sector goes hand in hand with industrial sector job growth rates of 6.5 percent per year growth in the non-traded sector. The 10 percent of cities in (recall, however, that the sector encompasses manufactur- which traded sector employment grew fastest from 2005 to ing plus construction, mining, and natural resources). After 2012 average 9.8 percent annual job growth in the sector, tumbling severely from 2005 to 2007 before climbing back, 6.6 percent outside the sector (that is, the non-traded sector), the sector’s growth rate in low-income cities appears to have and 8.1 percent economy-wide. By contrast, the bottom 90 converged with that of lower-middle-income and upper-mid- percent of cities average 2.0 percent annual job growth in dle-income cities at just above 2.0 percent annually. If such each sector respectively and citywide. Low- and middle-in- low job growth rates persist, the sector will indeed have lost come cities hoping to use exports to move up the value chain its traditional role as a sponge for excess labor, an engine of and develop their economies are not the only places where productivity, and a ladder of development. tradable sector growth differentiates competitive cities from their peers. Vibrancy in the traded sector appears to benefit The data also show that, globally, industrial job growth rates higher-income cities too: in cities with a GDP per capita above have held up slightly better in cities where manufacturing US$20,000, annual job growth averaged only 0.9 percent. is the dominant sector, averaging 3.9 percent from 2001 However, the 10 percent of cities experiencing the fastest job to 2003 and 2.9 percent from 2010 to 2012 (figure 2.8). In growth in the traded sector achieved, on average, 5.7 percent comparison, industrial sector job growth rates are low in annual growth in the traded sector and 5.3 percent growth cities dominated by high-end services, suggesting that the economy-wide. That is more than a fivefold increase in annual two sectors may have competing needs and accordingly that growth rates in what are typically mature economies. a natural sorting of activities occurs into cities specialized in one sector or the other. Primary cities narrowly outperform secondary cities globally for industrial sector job growth, but the two growth rates tend to move together. 17 Foreign direct investment FDI is spreading to more cities, but it remains highly con- centrated. In 2003, the top 10 percent of city destinations Foreign direct investment is frequently thought to both captured 71 percent of all FDI projects, 77 percent of all FDI signal and reinforce city competitiveness. Multinationals can jobs, and 78 percent of all FDI expenditures. By 2012, those typically select from a world of locations for their operations numbers had fallen to 67 percent, 67 percent, and 70 percent, and, therefore, tend to locate in cities that offer the right mix respectively. By comparison, the top 10 percent of cities con- of quality inputs, knowledge assets, infrastructure, supply centrated only 56 percent of GDP. chains, and market access they need to thrive—in short, competitive cities. At the same time, because of their pre- In absolute terms, the top destinations for FDI over the sumptive superior technology, management know-how, mar- period were the expected global centers of business, finance, keting prowess, and other proprietary assets, multinationals or production; trade entrepôts; and locales that have made are thought to offer special promise for locales wishing to FDI a pivotal part of their development model. Table 2.3 lists upgrade their local economies and accelerate development. the top 10 destinations for FDI in terms of cumulative capital However, the evidence is mixed on the extent to which pro- expenditure from 2003 to 2012. ductivity spillovers and technology transfer truly materialize from FDI and when.6 Nevertheless, because cities are eager Controlling for GDP yields a very different list of places where to attract FDI and are learning how to better integrate it into FDI factors most significantly in the local economy. Dividing their local economies, we examine its geography here. the total amount of FDI received over the 10-year period by GDP in the average year reveals cities—some expected, some To examine FDI performance at the city level, we were able unexpected—that punch well above their weight in terms to obtain data for 673 of our 750 cities from fDi Markets, a of FDI attraction. Table 2.4 lists the top 10 locales on this proprietary database maintained by the Financial Times that measure. Abuja (Nigeria) ranks first, drawing in more than aggregates data from press releases and other sources about a US$1 billion less FDI than its larger compatriot, Lagos, announced green field investments. The database neither but over a much smaller denominator. Factory Asia features claims to achieve universal coverage nor tracks the fate of prominently; cities from East Asia and Pacific populate nearly investments over time to revise its estimates. Sometimes an half the table. Tangier (Morocco) likely combines strong local announcement contains too little information for a project to fundamentals with relative stability and proximity to the be assigned to a city. In other cases, fDi Markets must impute large European market to rise to the top of the list. Among investment details using proprietary methods. Its cities may the better-known case studies in globalization, only Indian not always coincide with the boundaries of our OE cities cities, such as Hyderabad and Bangalore, and Vietnamese either. For these reasons, the database is best regarded as a cities, such as Da Nang, Hanoi, and Hai Phong register here. good-quality but not comprehensive record of FDI announce- ments. FDI has its largest influence in market towns and production centers. East Asia and Pacific, South Asia, and Sub-Saharan The three metrics contained in the database—number of Africa each contain 15 cities (or 22 percent of cities) in the projects, associated capital expenditure, and associated top 10 percent of all cities on this measure (figure 2.9, panel jobs—are flow variables based on new announcements each a). In contrast, the OECD is home to only one city with such year. The data are therefore lumpy, subject to volatility, and an FDI-oriented economy, Wrocław (Poland). Services hubs in highly cyclical. For that reason, much of the following analy- general constituted only 6 percent of the top echelon of per- sis looks at city FDI performance aggregated over the entire formers; production centers, conversely, constituted nearly period for which data were available (2003–12), because half (figure 2.9, panel b). Even on this measure, however, FDI city competitiveness should transcend economic cycles. The remains concentrated: the average city in the top 10 percent downside of this approach is that the world looked very on this measure received over 11 times more FDI per dollar of different in 2012 than it did in 2003—the dramatic fall of GDP than the average city in the bottom 90 percent. Cairo (Egypt) from the investor interest league tables being a case in point—and the method accordingly sacrifices some The economic development potential of FDI should not be recency. overstated, however. In fact, in the average city that received any FDI at all in 2012, foreign investors created only 1,400 jobs directly. That represented 0.1 percent of the average city’s employment base, or only a small fraction of the 2.0 percent net job growth that occurred on average in these cities. The truth remains that the majority of jobs, in every city, are still created by local, incumbent firms. 18 Table 2.3 Top 10 cities for FDI capital expenditures (cumulative 2003–12) FDI jobs, FDI projects, FDI capital expenditures, Rank City Country Region Typology 2003–12 2003–12 2003–12 (US$, millions) 1 Shanghai China East Asia and Pacific Production center 469,901 2,992 157,653 2 Singapore Singapore East Asia and Pacific Services hub 251,785 2,614 125,467 3 Beijing China East Asia and Pacific Production center 221,664 1,510 76,084 Organisation for Eco- United 4 London nomic Cooperation Services hub 100,789 2,346 72,827 Kingdom and Development United Arab Middle East and 5 Dubai Services hub 167,227 1,988 63,614 Emirates North Africa Hong 6 Kong China East Asia and Pacific Services hub 104,261 1,666 48,391 SAR Egypt, Arab Middle East and 7 Cairo Production center 52,054 210 40,268 Rep. North Africa Guang- 8 China East Asia and Pacific Production center 127,854 528 38,534 zhou 9 Bangalore India South Asia Production center 254,815 1,075 38,315 Suzhou, 10 China East Asia and Pacific Services hub 150,716 520 34,933 Jiangsu Source: fDi Markets, Financial Times Table 2.4 Top 10 cities for FDI inflows controlling for GDP (cumulative FDI 2003–12) FDI FDI capital expen- FDI capital expen- FDI jobs projects ditures 2003–12 ditures / average Rank City Country Region Typology 2003–12 2003–12 (US$, millions) GDP (US$) 1 Abuja Nigeria Sub-Saharan Africa Market town 12,744 16 4,393 8.21 2 Da Nang Vietnam East Asia and Pacific Market town 22,328 49 3,901 4.09 3 Ujung Indonesia East Asia and Pacific Market town 3,532 8 7,249 3.76 Pandang 4 Phnom Cambodia East Asia and Pacific Market town 15,065 96 4,820 3.29 Penh 5 Tangier Morocco Middle East and Production center 27,219 61 5,694 2.92 North Africa 6 Hyder- India South Asia Market town 136,323 431 21,428 2.46 abad 7 Bangalore India South Asia Production center 254,815 1,075 38,315 2.37 8 Vientiane Lao PDR East Asia and Pacific Market town 3,277 29 1,621 2.32 9 Hanoi Vietnam East Asia and Pacific Market town 83,940 346 18,442 2.02 10 Hai Phong Vietnam East Asia and Pacific Market town 31,874 57 4,279 1.95 Source: fDi Markets, Financial Times 19 Figure 2.9 Distribution of the top 10 percent of cities in terms of FDI, controlled for GDP By region By city economic structure % % Source: fDi Markets, Financial Times Note: See the discussion under industry pathways in section 3 of this paper for an explanation of this typology. 3. Development Pathways: How Do Income brackets may be City Economies Evolve over Time? useful for grouping cities The 46 cities in which according to their stage of GDP per capita fell This work stream hypothesizes development pathways fol- development, as is common- ly done for countries, but over the period provide lowed by cities and tests whether groupings of cities based on such brackets are static in a cautionary reminder characteristics such as population size, region of the world, stage of development, or industry specialization could be use- nature, and the spectrum it- that development is ful in diagnosing opportunities and constraints. Three path- self is discontinuous: a US$1 neither guaranteed ways based on the following characteristics were used: (a) city increase in GDP per capita nor a process of steady income level, hypothesizing that economic development is could result in a city’s reclas- advancement. a linear process through which cities move from low to high sification to a higher state income; (b) industry specialization, hypothesizing that cities, of development. For this as they develop, would first industrialize and then grow to reason, our analysis focuses specialize in high-end service industries; and (c) primacy or on the idea of development trajectories, classifying cities into population size within a national context, hypothesizing that groups on the basis of their annual average GDP per capita a city’s position within its national urban system should, all growth rates over the 12-year period in the OE data set. The else equal, remain stable as both develop. characteristics of cities growing at roughly similar rates— that is, on roughly similar trajectories—can then be studied Income pathway for similarities or differences. The most important development pathway for cities is To conduct this analysis, we divided the 750 cities in the OE arguably the progression from low-income to lower-middle, data set into five groups according to the rates of GDP per upper-middle, and finally high-income status. This pathway capita growth registered over the period. At the top was a is important because GDP per capita is still the most widely group of 135 cities that achieved GDP per capita growth rates accepted indicator of welfare and human development avail- above 10.0 percent on average each year—astonishing rates able. of economic development. Below them was another group of 133 cities that were on a slightly less rapid but still impressive The income pathway itself needs very little description: if we trajectory, with annual average GDP per capita growth rates assume that cities strive to achieve ever higher levels of GDP between 5.0 and 10.0 percent. In the middle was a cohort of per capita for their residents—again, a proxy for citizen wel- 229 cities that achieved steady average annual advances in fare and prosperity—then the relevant pathway is a simple GDP per capita of between 2.0 and 5.0 percent. Below that linear progression from a lower-income status to a higher-in- was a group of 207 cities that registered only modest but still come status. positive improvements in GDP per capita with growth rates below 2.0 percent. Finally, GDP per capita fell in 46 cities in 20 the sample set over the period. These cities, which had a nega- and three Peruvian cities—Arequipa, Lima, and Trujillo— tive annual average growth rate, provide a cautionary remind- represented Latin America and the Caribbean. Cities from all er that city development—at least as measured by output per across Europe and Central Asia landed in this group: Yerevan worker—is not always unidirectional. (Armenia), Sofia (Bulgaria), Riga (Latvia), Chisinau (Mol- dova), St. Petersburg and Vladivostok (Russia), Dushanbe The following subsections examine the cities in each group by (Tajikistan), Tashkent (Uzbekistan). No OECD cities passed region, income level in 2000, and industrial profile in 2000. the 5.0 percent threshold, however. City GDP per capita growth by region In the middle group of cities—those achieving 2.0 to 5.0 per- cent GDP per capita growth on average annually—we find the Powered almost exclusively by China’s rise, nearly all cities first OECD cities, a large contingent from Latin America and that achieved greater than 10 percent annual average GDP the Caribbean, and shrinking numbers from the fast-devel- per capita growth from 2000 to 2012 could be found in East oping regions of East Asia and Pacific and South Asia. In gen- Asia and Pacific (figure 3.1). The lone city from outside the re- eral, however, representation in this group was fairly broad gion was Baku (Azerbaijan), where GDP per capita increased based across regions and countries. The OECD cities encom- by 10.6 percent annually. passed a broad mix: resource-intensive Perth (Australia) and Aberdeen (United Kingdom); high-tech Portland Oregon and Indian cities constituted the largest national bloc in the San Jose (both in the United States); and Brisbane (Australia) second-fastest group of developers. Several East Asian and and Quebec (Canada) in the most developed countries. The Pacific cities outside China joined this group as well, includ- Republic of Korea’s cities, Poland’s cities, and other strongly ing Vientiane (Lao People’s Democratic Republic); Ulaan- performing cities in Eastern Europe, such as Prague (Czech baatar (Mongolia); and multiple cities in Indonesia, Malaysia, Republic), Talinn (Estonia), and Bratislava (Slovak Republic), and Vietnam. Nigerian cities dominated the contingent from completed the group. Several Russian cities and most Turkish Sub-Saharan Africa but were joined by Huambo and Luanda cities rank in this middle group and are joined by Tirana (Al- (Angola), N’Djamena (Chad), and Addis Ababa (Ethiopia)— bania), Sarajevo (Bosnia and Herzegovina), Zagreb (Croatia), three of the latter being national capitals. Marrakesh and and Belgrade (Serbia) in the Balkans; Astana (Kazakhstan) Meknes (Morocco) both made it into this group from the and Bishkek (Kyrgyz Republic) in Central Asia; and Dniprop- Middle East and North Africa, and Panama City (Panama) etrovsk, Kharkov, and Odessa (all in Ukraine). Wealthy Hong Figure 3.1 Annual average GDP per capita growth rates by region for 750 cities, 2000–12 Source: Oxford Economics Dataset Note: If data were not available for 2000, the first available year was used. 21 Kong SAR, China, and Singapore joined peers from across joined too. More troubling were the 12 cities from Sub-Saha- East Asia and Pacific. ran Africa in which development retrogressed on this mea- sure. These struggling cities included Yaoundé (Cameroon), The OECD contingent balloons in the bracket from 0.0 to 2.0 Bangui (Central African Republic), Abidjan (Côte d’Ivoire), percent annual average growth. Laggards from other regions Bamako (Mali), Niamey (Niger), Port Harcourt (Nigeria), and without the luxury of already high levels of GDP per capita Lomé (Togo). In the Middle East and North Africa, Aden and include Samarinda (Indonesia) and Kuching (Malaysia) from Sanaa (both in the Republic of Yemen) were accompanied— East Asia and Pacific, Srinagar (India) and Kathmandu (Ne- probably through a quirk of the data—by oil-rich United pal) from South Asia, and Skopje (former Yugoslav Republic Arab Emirates cities Abu Dhabi, Dubai, and Sharjah as well as of Macedonia) and several Ukrainian cities from Europe Kuwait City (Kuwait), where the shortcomings of our GDP- and Central Asia. Politically and economically struggling based measure become apparent. Unsurprisingly Port-au- Egypt populates most of the contingent from the Middle Prince (Haiti)—still reeling from the earthquake and a larger East and North Africa, alongside Mosul (Iraq), Casablanca institutional breakdown—landed here, too. São José dos (Morocco), and Sfax (Tunisia) as well as three Iranian cities. Campos in Brazil and Tampico and Tijuana from Mexico also Several cities of Sub-Saharan Africa remained mired here registered negative GDP per capita growth over the period. with near-stagnant development rates, including Cotonou (Benin), Brazzaville (Republic of Congo), Banjul (The Gambia), A few interesting patterns emerge to summarize. Nearly Conakry (Guineau), Nairobi (Kenya), Dakar (Senegal), and every East Asian and Pacific city managed to post annual Harare (Zimbabwe). South Africa’s cities joined them. The average GDP per capita growth rates over 2.0 percent—in second-largest contingent in this group comes from Latin most cases well over. Nearly all cities of South Asia and of America and the Caribbean, where many Brazilian and Mex- Europe and Central Asia, for their part, fell between 2.0 and ican cities eked out growth alongside numerous other cities 10.0 percent annual average growth. Cities of Sub-Saharan spread throughout the region. Africa appeared up and down the ranks of the distribution— too many too low, given their relatively low starting levels of At the bottom of the distribution are 46 cities in which GDP development. Cities of Latin America and the Caribbean and per capita actually fell over the 12-year period studied. The of the Middle East and North Africa found company in each bulk of these cities could be found in the OECD and were other in most brackets, whereas OECD cities—in large part badly hit by the 2009 financial crisis in Greece, Italy, Portu- because of high starting levels of GDP per capita—populated gal, Spain, and the United States. The Hague (Netherlands) the lower half of the distribution. 22 City GDP per capita growth by region, controlling The cities of the Middle East and North Africa appeared for country effects remarkably less vibrant by contrast. GDP per capita grew more slowly than it did at the national level in 70 percent Disentangling city GDP per capita growth from country GDP of the region’s cities—an illuminating finding in a socially per capita growth enables us to identify the outlier cities unstable region. Marrakesh stands apart with an annual whose rapid development trajectory stands apart against average GDP per capita growth 3.0 percentage points above backdrops of national economic growth or stagnation. that of Morocco. It is followed by Abu Dhabi and Sharjah in the United Arab Emirates and then Constantine in Algeria. As figure 3.2 shows, in East Asia and Pacific, nearly 90 per- In Latin America and the Caribbean, the OECD group, and cent of cities experienced faster GDP per capita growth than South Asia, GDP per capita increased slower than it did na- their national economies. Moreover, in almost half of East tionally in a majority of cities. In South Asia, the Indian cities Asian and Pacific cities GDP per capita rose annually above of Surat and Varanasi both achieved GDP per capita growth national rates by more than 2.0 percentage points on aver- rates more than 2.0 percentage points higher than that of age—a remarkable testament to the wealth-generating power India itself. of cities. Elsewhere in the world, only in Europe and Central Asia did a majority of cities similarly outpace their national economies on this measure. In Sub-Saharan Africa, as many cities outpaced their national economies in terms of GDP per capita growth as fell behind. Figure 3.2 Share of cities over-performing or underperforming their national economies in terms of average annual GDP per capita growth, by region, 2000–12 Overperforming by > 2% Overperforming by < 2% Underperforming by > -2% Underperforming by < -2% 0 10 20 30 40 50 60 70 80 90 100 Percent East Asia and Pacific Europe and Central Asia Sub-Saharan Africa Latin America and the Caribbean OECD South Asia Middle East and North Africa Source: Oxford Economics Dataset 23 City GDP per capita growth by income level industrialization begins. More clearly, though, the finding stands as a reminder that China—home to nearly every city As alluded to in the previous section and throughout this posting GDP per capita growth over 10.0 percent—contains technical note, a city’s starting point goes a long way in cities at all stages of development. explaining its economic performance vis-à-vis other cities globally. The poorer a city is, the higher its potential growth Low-income cities do account for the largest share of cities in rate tends to be; the richer a city is, the lower its potential the second-fastest growth group, with Indian and Nigerian growth rate will be. Figure 3.3 supports that rule broadly but cities well represented. Three high-income cities posted very shows that exceptions do exist. high average annual growth rates as well: Macao SAR (China), Almaty (Kazakhstan), and Perm (Russia). For this exercise, cities were slotted into income brackets on the basis of their GDP per capita in 2000. The income This finding suggests the existence of a global convergence or brackets were defined by the same cutoffs that the World catch-up story at the city level. In fact, the speed of city con- Bank Group’s official country classification scheme uses; vergence is faster than that of country convergence. This sug- these cutoffs are based on gross national income (GNI) per gests that cities lead economic growth and poverty reduction, capita. Although GDP and GNI are clearly different measures, a finding that is consistent with regional growth literature no equivalent classification scheme could be found for GDP (see Gennaioli and others 2014). A five-year lagged regression per capita, and applying the GNI cutoffs to GDP has produced model controlling for fixed effects is used to calculate city intuitive, reasonable, and nonarbitrary results. For the sake economic convergence rate. A conditional convergence rate of simplicity, income brackets as defined in 2012 were used to of 1.4 percent to 9.0 percent per year is observed for the 750 classify cities in 2000. largest cities in the world from 2000 to 2012. In other words, cities with a lower per capita GDP are catching up at a rate of A first striking finding is that more than two-thirds of the 1.4 to 9.0 percent per year.7 The unconditional convergence fastest-growing cities were lower-middle-income cities rather rate is also calculated at an interval of 1.9 to 4.5 percent per than low-income cities. This outcome may be because the year. process of development benefits from momentum, and cities very quickly graduate into lower-middle-income status once Figure 3.3: Annual average GDP per capita growth rates by starting income level for 750 cities (2000–12) Source: Oxford Economics Dataset Note: If data were not available for 2000, the first available year was used. 24 For the group of cities where GDP per capita grew between cutoffs over the period, falling from lower-middle-income to 2.0 and 5.0 percent on average annually between 2000 and low-income status. A comparatively large number of low- 2012, cities at all starting income levels could be found, er-middle-income cities were among these negative growers, which means an annual per capita GDP growth rate of 2.0 to too. In general, though, high-income cities dominated the 5.0 percent is commonly observed among a variety of cities, ranks of those struggling to maintain standards of living. from low income to high income (figure 3.3). City GDP per capita growth by income level, Among the 207 cities that eked by with slow but still positive controlling for country effects GDP per capita growth were only three low-income cities: Cotonou (Benin), Banjul (The Gambia), and Srinagar (India). Lower-middle-income cities are most likely to have expe- More than two-thirds of the cities falling into this group rienced faster GDP per capita growth than their countries. were high-income cities, attesting to the challenge of main- Figure 3.4 shows city GDP per capita growth outperformed taining high per capita growth rates at high starting levels national GDP per capita growth in 70 percent of the cities of income. Of course, the 2009 financial crisis that struck that were lower-middle income in 2000. China accounted for high-income cities (generally the same as OECD cities, as almost three-fifths of those cities. Outperforming the rate of mentioned previously) would also have contributed to these national development appears to be harder in high-income low growth rates. cities: only 2 percent did so by more than 2 percentage points per year, and it is the only income grouping in which fewer Evidence of the so-called poverty trap could be found in the than half outperformed national rates at all. By the same to- three cities—Bangui (Central African Republic), Bamako ken, high-income cities were less likely to lag far behind their (Mali), and Niamey (Niger)—that experienced negative GDP national economies in terms of GDP per capita growth, sug- per capita growth rates and were already low-income cities in gesting a stable, if slow and inertial, trajectory. Upper-mid- 2000. Lomé (Togo), too, had the inglorious distinction of be- dle-income cities, for their part, were most susceptible to ing the only city to regress to an earlier stage of development severe underperformance (GDP per capita growth rates more according to this measure and the World Bank Group’s official than 2 percentage points below the national rates). Figure 3.4 Share of cities by income bracket overperforming or underperforming their national economies in terms of average annual GDP per capita growth, 2000–12 Overperforming by > 2% Overperforming by < 2% Underperforming by > -2% Underperforming by < -2% 0 10 20 30 40 50 60 70 80 90 100 Percent High Income Upper Middle Income Lower Middle Income Low Income Source: Oxford Economics Dataset Note: If data were not available for 2000, the first available year was used. 25 City GDP per capita growth by industrial profile The high-income OECD cities that populate the lower rungs of the distribution also tend to specialize in high-end ser- Finally, we categorized cities by annual average GDP per cap- vices, which figure 3.5 makes clear. Causal interpretations of ita growth rates according to the industrial profile—the larg- this connection must be limited, however: a specialization in est sector in terms of share of GVA. In general, an orientation a certain sector does not cause a city’s GDP per capita growth toward industrial pursuits (which encompasses manufactur- rate to be what it is. Instead, sector specialization is more ing, energy, mining, and construction) was associated with likely to be associated with the level of development rather faster GDP per capita growth rates, whereas cities with lower than to be a driver of economic performance. As subsequent GDP per capita growth rates were more likely to have econo- sections of this paper discuss, another key development mies oriented toward high-end services. By contrast, cities in pathway for cities is that from basic services to industry and, which consumer (or basic) services were the dominant sector finally, to high-end services. Later sections expound on this tended to fall into the second- and third-fastest groups of idea in greater detail. growers—in part because of their tendency to be low- and lower-middle-income cities. Income traps Thanks to the dominance of manufacturing and construc- Income pathways can also be depicted by plotting cities’ tion in China’s economy, 95 percent of cities registering growth rates by their initial level of development (or GDP 10.0 percent or higher GDP per capita growth were indus- per capita). Figure 3.6 does this, plotting all cities by average trial cities. That number fell to 58 percent in the 5.0 to 10.0 GDP growth from 2002 to 2012 against GDP per capita at percent growth bracket. Cities with large consumer services the start of the decade, in 2002. The graph shows that growth sectors accounted for the second-largest contingent in this rates tend to decline as income levels rise in an economy. bracket. We posit that these cities’ growth may not be driven This slowdown takes place for a variety of reasons, including by this specialization, however, but rather that many cities the fact that richer societies at later development stages tend in the early stages of industrialization begin the process with to have fewer underused resources to marshal because the a dominant low-end services sector serving an increasingly formal sector already dominates, the urbanization process is urbanized population. Some of these fast-growing cities far along, and population levels stabilize. In addition, more specializing in high-end services are Beijing (China); Addis developed societies tend to operate at the technological fron- Ababa (Ethiopia); Bangalore Hyderabad, and Mumbai (India); tier and therefore have less capacity for growth that takes Riga (Latvia); Vilnius (Lithuania); and Panama City (Pana- advantage of technologies developed elsewhere to accelerate ma). Figure 3.5 Annual average GDP per capita growth rates by industrial profile for 750 cities, 2000–12 Source: Oxford Economics Dataset Note: If data were not available for 2000, the first available year was used. 26 the development process. In other words, easy efficiencies (Madagascar), and have mostly been exploited in wealthier cities, and growth Niamey (Niger), among A city may find itself in the rates regress to the rate of technological change. others, may be mired so-called middle-income in a poverty trap. The Maintaining growth rates high enough to keep the economy post-middle-income trap if its economy exhausts moving forward and up the income ladder is a priority—and trap could be defined the easy dividends of inte- worry—for many cities and countries. The frequently men- flexibly, and here we grating additional factors tioned middle-income trap refers to the symptom of growth posit that it begins just into the production process that slows before an economy becomes truly wealthy. Such after US$12,000 GDP more efficiently and fails to a slowdown may occur when an economy exhausts the easy per capita and extends dividends of integrating additional factors into the produc- cultivate a technology-driv- past US$20,000. By tion process more efficiently and fails to cultivate a technolo- these criteria, Porto en engine for the economy. gy-driven engine for the economy. Alegre and Rio de Ja- neiro (Brazil), Thessa- In figure 3.6, cities potentially afflicted by such a middle-in- loniki (Greece), and Szczecin (Poland), to name a few, may all come trap are the ones with low average annual growth find themselves in the post-middle-income trap, struggling to rates over the decade—possibly below 4.0 percent but more rekindle growth that prematurely petered out. stringently below 2.0 percent—and landing between approx- imately US$5,000 and US$12,000 in GDP per capita. By this At the same time, a number of upper-middle-income and definition, several Brazilian, Egyptian, and South African high-income cities in East Asia and Pacific, Europe and Cen- cities may be stuck in the middle-income trap as well as San tral Asia, and Latin America and the Caribbean enjoyed an- Salvador (El Salvador), Kingston (Jamaica), Asunción (Para- nual average GDP growth of more than 5.0 percent over the guay), and Caracas (República Bolivariana de Venezuela). decade. These stars—42 in total—include several Chinese cities as well as Mendoza (Argentina), Panama City (Panama), Cities may also find themselves in poverty traps, growing and Arequipa and Lima (Peru) in Latin America and Caribbe- very slowly even at extremely low levels of GDP per capita, an and Baku (Azerbaijan), Sofia (Bulgaria), Almaty (Kazakh- and post-middle-income traps, where city growth rates plum- stan), and Bucharest (Romania) in Europe and Central Asia. met before they approach the OECD average. By that defini- tion, Yaoundé (Cameroon), Bangui (Central African Republic), Abidjan (Côte d’Ivoire), Port-au-Prince (Haiti), Antananarivo Figure 3.6 Average GDP growth rates by GDP per capita in 2002, 2002–12 Source: Oxford Economics Dataset 27 Industry pathways cities specialize in a mix of basic and At early stages of development, Globally and measured broadly, the industrial composition of some higher-level a relatively large share of city city economies remained relatively stable over the 12-year pe- services. Once riod, suggesting that significant changes to the distribution industrialization value added in the consumer of economic activity across sectors in cities typically occurs takes hold, cities services sector is more a symp- over longer time horizons. Excluding China, where the in- take on a new tom of the underdevelopment dustrial sector dominates, the same sector accounted for the role as aggrega- of industry than it is the sign of largest share of GVA in 2012 as it did in 2000 in 80 percent tors of inputs a thriving consumer economy. of the cities in the data set. Of those cities that did change, and generators of Large consumer services sectors just below half underwent a transition from specializing in economies of scale industry to specializing in high-end services. Sixteen mostly and agglomeration may also appear when the pro- South Asian cities transitioned directly from basic consum- forces. Eventually, cess of urbanization precedes er services into high-end services, bypassing the process of as development that of industrialization. industrialization for the time being (see figure 3.7). These continues, econ- cities could be considered sector leapfroggers, although their omies transition continued progression up the economic development ladder into higher-level services functions while the consumer without a strong industrial base may not be guaranteed. services sector falls into an equilibrium size to support more dominant traded sectors. Accordingly, the size of the consum- Note: Blue blocks represent all cities with industry as the er sector appears to be a function of the size of other sectors largest sector in 2012, gray blocks are those where high-end at all stages of development. At early stages of development, a services are the largest sector, and red blocks represent cities relatively large share of city value added in the consumer ser- with consumer services as the largest sector. vices sector is more a symptom of the underdevelopment of Theory would suggest that cities embark on a slightly differ- industry than it is the sign of a thriving consumer economy. ent development process than do countries. Cities tend to by- Large consumer services sectors may also appear when the pass the earliest agrarian stages of development because they process of urbanization precedes that of industrialization. are, by nature, marketplaces for the exchange of goods and services. Accordingly, at the earliest stages of development, Figure 3.7 City sector pathways from 2000 to 2012 In contrast to income growth path, stability in aggregate-level sector composition in cities appears to hold over the medium-long (12 year) period studied 2000 2012 118 Industry Industry 165 cities 4 129 cities 7 10 Consumer services Consumer services 32 52 cities 49 cities 7 37 High-end services 16 High-end services 185 cities 224 cities 171 Sample set excludes cities from Argentina, China, Malaysia, Morocco, Peru, Russia, Serbia, South Africa, Turkey, United Arab Emirates, Ukraine, United States, Venezuela because of data availability. Source: Oxford Economics Dataset Note: Line width is only indicative. 28 As figures 3.8 and 3.9 show, industrialization appears to take category as productive centers. Finally, once GDP per capita off very quickly at the lowest levels of development, where it surpasses the US$20,000 level, high-end services take over remains a dominant force, on average, well into upper-mid- as the largest sector of the economy and presumptive driver dle-income status. At the city level, industry’s share of GVA of growth and wealth creation. In these hubs of creative appears to peak around US$11,000 GDP per capita. The tran- services, innovation-stimulating Jacobian externalities likely sition from industry to high-end services then takes place increase in importance. gradually as countries grow richer. As an aside, the pronounced differences between figures Figure 3.8 also implies that, structurally speaking, economic 3.8 and 3.9 (the latter incorporates China’s 150 cities in the development is a progression toward high-end services— sample) may lend further support to Rodrik’s proposition even if industry may be development’s engine. (The figure that the historical pathway to prosperity via industrializa- sorts cities according to ascending GDP per capita and places tion is closing for low-income cities today. This closure may them into bands of 12 cities. It then stacks each band’s be in part because of changes to the labor intensity of the column according to the average size of each sector of the manufacturing sector worldwide but also because of the economy, producing the picture shown.) sheer scale of China’s manufacturing enterprise. However, we have no historical period with which to compare the On the basis of figure 3.9, we posit that cities can general- data in figure 3.9, and we cannot assess whether crowding ly be classified into three different categories according to out is taking place without knowing whether the amount their stage of development. Until GDP per capita reaches of industrial activity the world economy can support at any approximately US$2,500 per year, cities fundamentally given time is finite. One could therefore arrive at two very serve as market towns in the classic sense; they are centers different conclusions on the basis of these figures. One poten- of exchange for people trading their wares on a small scale tial conclusion is that the graph including China reflects the and from a surrounding hinterland. At this stage, consumer global steady state—one in which Chinese cities are winning services dominate the economy. Then, as GDP per capita rises the biggest pieces of the global industrial pie. The other to about US$20,000 per year, industry’s share of the econo- possible conclusion is that China is an anomaly—indeed, it my also begins to rise and remains dominant. At this stage, is worth recalling that construction, in addition to mining, is cities serve as industrial agglomerations and benefit from included in the “industrial” aggregate, a sector that has seen traditional Marshallian externalities.8 We label cities in this very large booms in China—and its cities buck the normal trend observed in figure 3.8. The truth, of course, may also lie somewhere in between. 29 Figure 3.8 The changing distribution of GVA across sectors as GDP per capita rises, 2012, excluding China Percent $2,000 $5,000 $12,000 $40,000 100 90 High-end services 80 Sector share of total gobal value added (%) 70 60 Public services 50 40 Consumer services 30 Agriculture 20 Industry and mining 10 0 Band of 12 cities ordered from lowest to highest GDP per capita Source: Oxford Economics Dataset Figure 3.9 The changing distribution of GVA across sectors as GDP per capita rises, 2012, including China Percent $2,000 $5,000 $12,000 $40,000 100 90 High-end services 80 Sector share of total gobal value added (%) 70 60 Public services 50 40 Consumer services 30 Agriculture 20 Industry and mining 10 0 Band of 12 cities ordered from lowest to highest GDP per capita 30 Source: Oxford Economics Dataset Size pathways Figure 3.10 Cities increasing and decreasing their shares of national population worldwide, 2000–12 Stability reigns at the top of national urban hierarchies (see figure 3.10). Over the 12-year period studied, no city was Point change in city share deposed from the top spot in its national hierarchy in terms of national population (%) of GDP, although five cities that were prime in terms of GDP 12 - rose to become prime in terms of population as well, dis- placing originally larger but less productive cities. In Bolivia, 8- La Paz’s population fell behind that of Santa Cruz, just as Lilongwe’s population overtook that of Blantyre in Malawi. In the OECD, Milan advanced past Naples to become Italy’s 4- Increasing share of most populous city, and Amsterdam displaced Rotterdam country population atop the Dutch urban hierarchy. Vietnam’s Ho Chi Minh City 0- surpassed Hanoi in East Asia and Pacific. Within large countries with multiple large cities, substan- Decreasing share of -4 - country population tial changes in position were rare but not unheard of. In the United States, cities such as Detroit and New Orleans fell five -8 - places in population rank just as Austin climbed five plac- es and Las Vegas and Orlando climbed six places. In terms Source: Oxford Economics Dataset of GDP, Portland, Charlotte, North Carolina; and Austin, Texas all moved significantly up in the rankings. In Nigeria, Onitsha rose eight places in terms of population. Of 72 major cities in India, 17 substantially (defined as a movement of at Figure 3.11 Cities increasing and decreasing their least five places in either direction) changed their position shares of national output worldwide, 2000–12 in the national hierarchy in terms of population, and 25 did in terms of GDP. In China, fully one-third of 150 large cities Point change in city share of national rose or fell significantly in the population rankings. In terms gross domestic product (%) of GDP, nearly 60 percent did. 12 - The same general distribution holds for changes in city share 8- of national GDP (see figure 3.11). Of the 553 cities for which data were available, 74 percent saw their share of national GDP increase over the period, 20 percent saw their share 4- Increasing share of decrease, and 6 percent saw their share remain unchanged. country population More turbulence was found in the distribution of national 0- output among cities than in population. Only 55 percent of cities saw their share of national output remain stable, defined as changing less than 0.2 percentage points in either -4 - direction, compared with 71 percent in terms of population. Decreasing share of country population -8 - Source: Oxford Economics Dataset 31 The share of national population increased in nearly 80 per- Regionally, nearly every cent of cities in rapidly urbanizing Sub-Saharan Africa from city in East Asia and Pacific, Whereas all city 2000 to 2012. By contrast, in more stable Europe and Central Europe and Central Asia, and boats rise together in Asia, only 50 percent of cities garnered a larger share of their Sub-Saharan Africa increased urbanizing countries, country’s population by 2012 (figure 3.12). Dynamic East its share of national GDP Asia and Pacific, for its part, saw significant jockeying among from 2000 to 2012. By con- industry specializa- cities, with as many rising in population share as falling or trast, in the Middle East and tions appear to drive remaining the same. South Asia stands out for a large cohort North Africa and the OECD, a city’s climb or fall in of cities that neither increased nor decreased their share of as many cities saw their share hierarchies of devel- country population over 12 years and for the smallest cohort of national output decline as oped countries. of cities ceding ground. Here a change is any increase or de- saw it increase. This differen- crease in country population share of 0.01 percentage points tial suggests that the relevant or greater. growth drivers may operate at the national level in the for- mer group of countries undergoing increasing urbanization broadly, but then operate at the city or particular industry level in countries with more established urban hierarchies, thus leading to shifts in a city’s share of output as cities rise and fall with the fate of their base industries. Figure 3.12 Direction of change in city’s gain in national population and output as urbanization occurs, by region Direction of change in city share of Direction of change in city share of national population national gross domestic product Share of cities in region (%) Share of cities in region (%) Urbanization 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 rate (%) Sub-Saharan Africa Sub-Saharan Africa 37.2 South Asia South Asia 31.2 Middle East and North Africa Middle East and North Africa 61.5 Latin America and Caribbean Latin America and Caribbean 81.3 OECD OECD 81.0 Europe and Central Asia Europe and Central Asia 65.5 East Asia and Pacific East Asia and Pacific 49.7 Direction of Change in City Share of National GDP Source: Oxford Economics Dataset Rising Unchanged Falling Europe and Central Asia Sub-Saharan Africa East Asia and Pacific South Asia Latin America and Caribbean Middle East and North Africa OECD 32 4. Predictors of City Competitiveness Stocktaking of city indexes A critical value added of this paper is to take stock of and Working definition of city competitiveness assess how the various city competitiveness indexes on the City competitiveness has been defined for the purposes of this market can be most useful for policy. The team therefore project as follows: a competitive city is a city that successfully inventoried popular indexes covering themes ranging from facilitates its firms and industries to create jobs, raise produc- competitiveness to livability, sustainability, and infrastruc- tivity, and increase the incomes of citizens over time. ture. These indexes vary significantly in terms of methodolo- gy and coverage. Several adopt opaque “black box” approaches The analytics team operationalized city competitiveness with to ranking cities that do not render them fit for statistical concrete and measurable economic outcomes. Although many analysis, whereas others show a clear bias toward only a few different indicators have merit, and each captures a different regions of the world, typically OECD countries. In total, more facet of competitiveness, we have adopted four on the basis of than 20 different indexes were evaluated according to the generalizability and availability: robustness of their methodologies as well as their coverage, and six were chosen for folding into the final data set (see • Annual output growth table 4.1). • Annual employment growth • Labor productivity (GVA per worker) Other data sets used at various points in this analysis include • Household disposable income the World Bank Group’s Doing Business Indicators (DBI), the Chinese Academy of Social Science’s Global Urban Competi- These outcomes allow us to explore a city’s ability to add tiveness Report for patent data, and the International Mone- value, grow jobs, raise productivity, and increase residents’ in- tary Fund’s Government Finance Statistics Yearbook (GFSY). comes—preoccupations of city leaders across the world—and the factors that are associated with such an ability. Table 4.1 Popular City Indexes Evaluated for Methodological Robustness and Coverage Evaluated and included in global correlation analysis Evaluated and not included in global correlation analysis AT Kearney Global Cities Index (2008, 2010, 2012) (65 cities) 2thinknow Consulting: Innovation Cities Index Economist Intelligence Unit City Competitiveness Hotspots BusinessWeek ’s America’s Best Cities (2012) (116 cities) Cisco Broadband Quality Index Economist Intelligence Unit Liveability Index (2012) (68 cities) fDi Intelligence: Cities of the Future Global Financial Centers Index (2006–12) (62 cities) Forbes Best Places for Business and Careers Mercer Quality of Life Index (2010, 2012) (46 cities) IDB Emerging and Sustainable Cities Initiative UN-Habitat City Prosperity Index (2012) (70 cities) Mercer Infrastructure Index Milken Institute’s Best-Performing Cities Monocle Quality of Life Survey NUMBEO Cost of Living and Quality of Life Indexes PWC Cities of Opportunity Siemens Green City Index Urban Sustainability Index World Knowledge Competitiveness Index 33 How well do city indexes predict competitiveness tions closer to the center are more strongly negative, just as outcomes? correlations toward the outside of the chart are more strongly positive.) The negative correlations should in many cases be Ultimately the team chose six indexes to test for their pre- expected: they imply slower rates of growth at higher rank- dictive power. The selection was based on transparency, data ings. If rankings are associated—as is likely—with stage of accessibility, coverage, and popular use. To assess whether development, then relatively slow growth rates in developed an index was a strong predictor of city competitiveness, we cities make for an inverse relationship. The Global Financial ran a straightforward pair wise correlation analysis between Centers Index, for its part, positively predicted every compet- the index’s 2012 rankings and the four primary performance itiveness outcome in 2012. outcomes in 2012. In general, no city competitiveness index can claim a univer- Globally, the indexes proved to be far better predictors of sal ability to predict the economic performance of the world’s the level of city development than predictors of short-term cities, especially in the short term. A gap in the market seems economic performance (see figure 4.1). The Economist to exist for indexes to capture cities’ growth potential. Intelligence Unit (EIU) Hotspots and Liveability surveys, AT Kearney’s Global Cities survey, and the UN-Habitat City Some indexes perform better in samples narrowed by either Prosperity Index are all highly correlated with both city labor region or city typology. The EIU Competitiveness Index, for productivity and household disposable income—two com- example, predicts outcomes better in East Asian and Pacific petitiveness outcomes that vary with overall level of develop- and OECD cities than it does in Latin American and Carib- ment but remain quite stable from year to year. Performance bean or South Asian cities and better in high-income cities on these same indexes was actually negatively correlated than in low-income ones. The AT Kearney Global Cities Index with GDP and employment growth from 2011 to 2012— performs better with primary cities, and the Mercer Quality more strongly so for GDP than for employment. (Negative of Life Index does better for secondary cities. correlations can be found at the center of the radar; correla- Figure 4.1 Correlations among city index rankings and city performance outcomes, 2012 Correlation -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 Labor productivity Labor Productivity EIU City Hot Spots Disposable Income per Capita Disposable income per capita Employment Growth GDP Growth UN-Habitat City Prosperity Index Job growth GDP growth EIU Liveability Index A.T. Kearney Global Cities Index Global Financial Centers Index Mercer Quality of Life Index Source: Oxford Economics Dataset and Various Indices Note: EIU = Economist Intelligence Unit. 34 City competitiveness correlates: Factor analysis Factor analysis: Global results on the determinants of city competitiveness Indexes are by nature composite measures of multiple factors. As a result, the information they provide is of only limited Each determinant can be represented by multiple indicators use to policy makers, who tend to see the world through or subcomponents of indexes. We used factor analysis meth- discrete issues, subject matters, and areas of expertise. odology to consolidate multiple related variables into a single Accordingly, we conducted a second correlation analysis to statistical entity. We then ran pair wise correlation analysis assess the relationship that particular subcomponents of across factors and outcomes to ascertain how the relationship indexes, as well as additional variables from the World Bank’s among determinants and outcomes varies by the city’s stage DBI database (with data for 2005 through 2012), have with of development. The stage of development is defined accord- city competitiveness outcomes. These subcomponents and ing to the typology on the basis of stage of industrialization measures were bucketed into six conceptual determinants of (discussed under industry pathways in section 3 of this city competitiveness: institutions and regulations, physical paper) that classifies cities according to their GDP per capita infrastructure, social infrastructure, human capital, innova- levels as market towns, production centers, or creative and tion, and enterprise support and finance (see table 4.2).9 financial services hubs. Figure 4.2 summarizes the observed relationships. Only statistically significant associations at We use the term determinant cautiously. Determinants are the 10 percent level are shown. The analysis incorporates meant to capture the variables we expect to be associated observations spanning several years because the historical with positive economic outcomes and city competitiveness. coverage of each data set used varies. We decided to retain However, any relationships identified here are by no means the full sample to incorporate the largest possible number of deterministic. They are associations and correlations and observations. should be interpreted only as such. Table 4.2 Determinants and proxies of city competitiveness Institutions and regulations Skills and innovation Ease of Doing Business Index (Doing Business Indicators) Human capital Education Index (EIU Livability) Human Capital (EIU Competitive Cities) Innovation Number of Patents by City (Global Urban Competitiveness Report) Infrastructure and land Enterprise support and fnance Physical infrastructure Financial Maturity Index (EIU Competitive Cities) Physical Capital (EIU Competitive Cities) Private Credit Bureau Coverage (Doing Business Indicators) Infrastructure Index (UN-Habitat City Prosperity Index) Cost of Electricity (Doing Business Indicators) Social infrastructure Social and Cultural Capital (EIU Competitive Cities) Healthcare Index (EIU Livability) Quality of Life Index (UN-Habitat City Prosperity Index) Source: Competitive cities for jobs and growth: what, who, and how, 2015, The World Bank. Note: EIU = Economist Intelligence Unit. 35 We find that the correlates of city competitiveness vary de- • Innovation is associated with labor productivity at higher pending on income level and industrial structure as follows: income levels and economic structures. • Institutions and regulations matter at most levels of in- • Human capital is positively associated with household come and economic structures. disposable income, labor productivity, and job growth, but only after cities reach high-income status dominated • Physical infrastructure is positively associated with job by high-end services sectors. growth in market towns and labor productivity in pro- duction centers. Our results lead to an intriguing hypothesis, coherent with findings in existing literature (Moretti 2004; Samad, Loza- • Social infrastructure appears to be important for income no-Gracia, and Panman 2012; Shapiro 2006; World Bank growth and labor productivity only in relatively well-de- 2009; World Bank and DRC 2014): cities can use a sequence veloped cities. of interventions. The building blocks of competitiveness— institutions and social and basic physical infrastructure at • Financial infrastructure appears to be important for lower incomes, then innovation capacity—apparently can labor productivity and employment growth in relative- be sequenced in priority to build the human capital base ly well-developed cities. (We did not find data sets or required to compete, grow, and prosper as a high-income city. indexes that cover the provision of enterprise support In prioritizing these interventions, policy makers should programs and, thus, could not include this factor in our keep in mind the main industrial structure of the city and its global analysis.) competitive advantages. Figure 4.2 Statistically significant correlations among factors of determinants and competitiveness outcomes by city type Market Towns Market Towns Creative Financial Centers (US$20,000 GDP per capita) Disp. Labor GDP Emp. Disp. Labor GDP Emp. Disp. Labor GDP Emp. income prod. growth growth income prod. growth growth income prod. growth growth Institutions and regulation Ease of doing business index (DB) Physical Infrastructure Physical infrastructure (EIU); infrastructure index (UN); cost of electricity (DB) Social infrastructure Social and cultural capital (EIU); healthcare (EIUL); quality of life (UN) Financial infrastructure Private credit bureau coverage (DB); financial maturity (EIU) Human capital Human capital (EIU); education (EIUL) Innovation Number of patents (GUC) Positive statistically significant correlation at the 10 percent level Negative statistically significant correlation at the 10 percent level Source: Oxford Economics Dataset and Various Indices Note: DB = Doing Business Indicators; EIU = Economist Intelligence Unit City Competitiveness Hotspots; EIUL = Economist Intelligence Unit Liveability Index; GUC = Global Urban Competitiveness Report; UN = UN-Habitat City Prosperity Index; Disp income - Disposable income; Labor prod = Labor productivity; GDP = gross domestic product; Emp. growth = employment growth. 36 Breaking the factors down into their component parts, as omies. Both measures of human capital are associated with figure 4.3 does, enables us to isolate in even greater detail job growth and productivity in services hubs, as are mature the determinants that appear to influence competitiveness financial markets. Together, these findings suggest that basic outcomes. (physical) and more advanced (human and financial) forms of capital may become increasingly complementary as cities Labor productivity appears to benefit most directly from develop. Social infrastructure, for its part, appears to be high- improvements on various determinants, especially in pro- ly correlated to living standards and development levels but duction centers and services hubs. Productivity and income exhibits no significant connection to growth rates. levels in market towns are correlated with lower costs of electricity and more developed financial markets. In produc- Predictors of city competitiveness by region and tion centers, electricity costs, physical capital, and human city type capital are all positively associated with output growth rates, whereas productivity appears to benefit from improvements We also carried out pair wise correlation analyses between on a number of different fronts. the indexes and competitiveness outcomes by region and types of cities. Physical infrastructure and human capital may become increasingly complementary as income levels rise. Lower Generally across regions and types, the finding holds that electricity costs are associated with higher job growth, out- city indexes are good at predicting a city’s overall level of put growth, and productivity levels even in higher-income development but are less useful at explaining differences in services hubs, underscoring the continued saliency of the performance among cities at the same stage of development. basic microeconomics of production even in services econ- Figure 4.3 Statistically significant correlations among determinants and competitiveness outcomes by city type Market Towns Market Towns Creative Financial Centers (US$20,000 GDP per capita) Disp. Labor GDP Emp. Disp. Labor GDP Emp. Disp. Labor GDP Emp. income prod. growth growth income prod. growth growth income prod. growth growth Institutions and regulation Ease of doing business index (DB) Physical Infrastructure Physical infrastructure (EIU) Infrastructure index (UN) Cost of electricity (DB) Social infrastructure Social and cultural capital (EIU) Healthcare (EIUL) Quality of life (UN) Financial infrastructure Private credit bureau coverage (DB) Financial maturity (EIU) Human capital Human capital (EIU) Education (EIUL) Innovation Number of patents (GUC) Positive statistically significant correlation at the 10 percent level Negative statistically significant correlation at the 10 percent level Source: Oxford Economics Dataset and Various Indices Note: DBI = Doing Business Indicators; EIU = Economist Intelligence Unit City Competitiveness Hotspots; EIUL = Economist Intelligence Unit Liveability In- dex; GUC = Global Urban Competitiveness Report; UN = UN-Habitat City Prosperity Index; Disp income - Disposable income; Labor prod = Labor productivity; GDP = gross domestic product; Emp. growth = employment growth. 37 Accordingly, factors are consistently better at predicting labor Figures 4.4, 4.5, and 4.6 illustrate these perspectives with productivity and disposable income levels than they are at actual data. Figure 4.4 explores the relationship between predicting the shorter-term changes in output or employ- the EIU’s financial maturity measure and competitiveness ment. With smaller sample sizes, even fewer correlation outcomes across World Bank Group regions. One can see that coefficients are statistically significant. Those caveats aside, the measure is positively correlated with labor productivity correlation analysis by city type constitutes a sound starting and disposable income per capita across all regions and with point for attempting to explain the variation in competitive- GDP growth in the Middle East and North Africa and Europe ness outcomes across cities. and Central Asia. One can conduct the narrower correlation analysis from one Figure 4.5 takes only East Asian and Pacific cities into of three different perspectives: that of a region, that of a account to show that across a set of indicators representing determinant, or that of a factor. For example, one can try to each determinant, institutional strength and social infra- uncover what factor is most strongly correlated with compet- structure are positively associated with job growth in the itiveness outcomes in groups of different cities. Alternatively, region. Observed relationships with GDP growth, in contrast, one could approach the analysis by asking where and for are all counterintuitively negative—hence suggesting that which outcome a particular determinant seems to matter. Or the indicator may actually be assessing something that is in- finally, a productivity scholar, for example, may wish to know versely related to the factors driving growth in this particular how the relationship between a set of determinants and labor context. Once again, associations with labor productivity and productivity differs across regions. disposable income levels are stronger and more positive. Figure 4.4 Example correlation analysis output between financial maturity and competitiveness outcomes, by region Europe and Central Asia Middle East and North Africa Disposable income per capita Labor productivity Sub-Saharan Africa Job growth East Asia Pacific GDP growth OECD South Asia Latin America and the Caribbean Source: Oxford Economics Dataset and Economist Intelligence Unit Figure 4.5 Example correlation analysis output between competitiveness determinants and outcomes for East Asian and Pacific cities Institutions Physical infrastructure Human capital Disposable income per capita Labor productivity Agglomeration Job growth Financial infrastructure GDP growth Social infrastructure Environment 38 Source: Oxford Economics Dataset and Various Indices Figure 4.6 displays correlations between the full range of How does the “mayor’s wedge” relate to city competitiveness determinants and labor productivity by re- competitiveness? gion. The example clearly shows that institutional strength is more highly correlated with labor productivity levels in cities Of particular interest is the idea of the “mayor’s wedge”—that of Sub-Saharan Africa (and less highly correlated in cities of is, the capacity, scope, and autonomy for relevant policy mak- Europe and Central Asia) than in cities of any other region, ing at the city level—and its relationship with city competi- just as social infrastructure is most strongly associated with tiveness outcomes: GDP and job growth, disposable incomes, higher labor productivity in cities of South Asia and least and labor productivity. strongly associated in cities of Europe and Central Asia. In cities of Europe and Central Asia, physical infrastructure and We explore the mayor’s wedge relationship with city competi- financial infrastructure are most closely associated with higher tiveness using an econometric model. We obtained data from levels of labor productivity. A chart such as this example could the International Monetary Fund’s GFSY database, which help scholars from different regions identify pertinent and provides information on spending at all levels of government promising issues for their competitiveness inquiries. in countries outside Asia. Coverage of this data set expand- ed with each year, beginning with 100 cities in the sample Because these are only correlations, and correlations reveal in 2000 and ending with 286 by 2012. OE data on social associations but offer little insight into explanatory factors, infrastructure and agglomeration, which had full coverage, later stages of this project may conduct a more comprehen- were also used to introduce a series of explanatory variables sive multivariate regression analysis to complement this base. covering additional competitiveness determinants. We performed a multivariate regression analysis to test the ability of two components of the wedge—scope and financial autonomy—to explain competitiveness outcomes. The ordi- Figure 4.6 Example correlation analysis output between labor productivity levels and competitiveness determinants, by region Physical infrastructure Financial infrastructure Human capital Social infrastructure Institutions Agglomeration Environment Source: Oxford Economics Dataset and Various Indices 39 nary least squares model controlled for city and year fixed interpreted as follows: a 1 percentage point increase in the effects to render a more robust result and reads as follows: local share of total government spending (a proxy we used for measuring the scope of a mayor’s administrative remit) (4.1) City Competitiveness Outcomesct=αc+βt+ is associated with a commensurate 0.132 percentage point δMayor’ sWedgect+xct+ εct, increase in job growth and a 0.179 percent decrease in house- hold disposable income. When financial autonomy is used Where αc and βt are city and year fixed effects, respectively, as a proxy for a mayor’s administrative remit, the regression and xct denotes the determinants of city competitiveness results show that a 1 percentage point increase in the share introduced as controls. The dependent variable measures of a city’s revenues raised locally is associated with a 0.174 competitiveness outcomes for a given city (denoted by the percentage point decrease in GDP growth, a 0.115 percent de- subscript c) at a given point in time (denoted by the subscript crease in labor productivity, and a 0.263 percent decrease in t). Similarly, the explanatory variable also varies by a given household disposable income. Explanations of these regres- city and year. The numerical coefficients are shown in table sion results are presented in the following section. 4.3 and depicted visually in figure 4.7. The results should be Table 4.3 Regression results from mayor’s wedge model Labor Disposable City competitiveness outcomes GDP growth Job growth productivity income Mayor’s wedge: scope 0.000335 0.00132*** −0.000337 −0.00179* (0.000827) (0.000405) (0.000643) (0.00101) −0.00174** −0.000629 −0.00115* −0.00263** Mayor’s wedge: financial autonomy (0.000809) (0.000406) (0.000656) (0.00113) Constant 10.10*** 6.425*** 3.521*** 3.431*** (0.322) (0.217) (0.202) (0.397) Observations 1,897 1,897 1,897 1,841 R-squared 0.998 0.998 0.997 0.992 Year fixed effects Yes Yes Yes Yes City fixed effects Yes Yes Yes Yes Source: authors’ own analysis; dataset used is authors combining oxford economics dataset and various indices Note: We also include a number of controls, such as physical infrastructure provisions: number of nurses, doctors, and teachers as well city Gini index to cap- ture the level of inequality. However, because they were not the focus of this regression exercise, we highlight only the mayor’s wedge results here. Significance level: * = 10 percent, ** = 5 percent, *** = 1 percent. Figure 4.7 Size, direction, and significance of mayor’s wedge coefficients from global regression analysis Scope * Disposable income per capita * Labor productivity Job growth GDP growth Autonomy * * Statistically significant correlation at the 10% level * * Correlation coefficient (% change) Source: authors’ own analysis; dataset used is authors combining oxford economics dataset and various indices 40 Significance level: * = 10 percent. Findings from the mayor’s wedge analysis: Scope sectors bloated with cronies may actually be associated with and autonomy improved economic outcomes. Complicating matters further, proxies of public sector capacity such as public sector output The analysis found that delegating more public functions to per worker were negatively associated with outcomes globally. mayors—that is, expanding the scope of a mayor’s adminis- trative remit, proxied by the local share of total government To try to get around these global issues, we explored the per- spending—is positively associated with higher levels of formance of proxies for capacity in two institutional contexts: employment in cities but negatively associated with income. China and the European Union. In China, we took advantage Financial autonomy—the share of local government revenues of a recent policy change that upgraded particular counties raised locally and not transferred by the central govern- to city status in an effort to better understand the economic ment—is negatively associated with all competitiveness out- effect of an enlarged administrative remit. We found that comes except job growth. This finding suggests that, globally, expanded scope alone was not sufficient for improving firm- cities perform better with stable revenue streams from the and city-level economic outcomes. Instead, only high-capacity central government. cities—defined here as those with (a) a larger share of public employees paid through public finance out of the total pop- Findings from the mayor’s wedge analysis: ulation (proxying for adequate and institutionalized human capital in the public sector) and (b) higher tax extraction Capacity capability, as defined by local tax revenues normalized by Another important aspect of the efficacy of local govern- local GDP—appear to succeed in translating an expanded ments is their ability to design and implement sound policies remit into improved competitiveness outcomes (see Zhu and in the spaces that fall within their remit—in short, their Mukim 2015). The other exercise took advantage of public capacity. Unfortunately, attempts to directly measure local employee productivity data from the European Union to find government capacity—the ability of a government to execute that cities with larger administrative scopes and high levels and implement policy efficiently and capably—accurately and of public sector productivity perform better on measures of systematically across countries met methodological road- job growth. Using the size of the public sector as a proxy for blocks. The size of the public sector, a proxy for inputs, was capacity, however, reveals a negative association with GDP, strongly positively correlated with outcomes but is fraught jobs, and income growth. This finding stands in contrast to with complications ranging from issues of endogeneity to the China finding. perverse conclusions, such as the implication that large public 41 Variation in the effect of the mayor’s wedge by The positive association between mayor’s scope and job region and city type growth observed at the global level extends only to OECD cities (with statistical significance) at the regional level, We then ran the regression by region and city typology and where it is also positively and significantly correlated with found that the nature and effect of the mayor’s wedge appears GDP growth. to vary significantly across regions (see figure 4.8). The may- or’s wedge appears to have better explanatory power in OECD Sub-Saharan African and European and Central Asian cities and Middle Eastern and North African cities than it does in drive the negative relationship observed between autonomy Latin American and Caribbean or Sub-Saharan African cities, and job growth globally. GDP growth in the Middle East and where we found few statistically significant relationships. North Africa and job growth in cities of the Middle East and North Africa and the OECD, in contrast, appear to benefit from local government autonomy. Figure 4.8: Size, direction, and significance of mayor’s wedge coefficients from regional regression analysis Gross Domestic Product Autonomy Jobs Scope Middle East and North Africa * OECD * OECD * Middle East and North Africa * * Latin America and Caribbean Latin America and Caribbean Sub-Saharan Africa Europe and Central Asia * * Europe and Central Asia Sub-Saharan Africa * Correlation coefficient (% change) Correlation coefficient (% change) * Statistically significant correlation at the 10% level Source: Authors’ own analysis; dataset used is authors combining Oxford Economics dataset and various indices 42 Cutting the data by city primacy and size instead of by region GDP and job growth in industrial cities. We expect industrial reveals a different set of findings (figure 4.9). The scope of cities to vary more in their economic specialization than cit- a mayor’s remit is positively associated with job growth at ies oriented to either consumer or high-end services sectors, all city sizes, but the relationship is stronger in small and given what we know about the economics of agglomeration. secondary cities than it is in large or primary ones. Fiscal The logic is that local leaders may therefore be better attuned autonomy is negatively associated with job growth and GDP to the particular needs of their specialized regional econo- growth in small and secondary cities but shows little relation- mies. ship in large or primary ones.10 Together these findings suggest—perhaps intuitively—that These findings suggest that small and secondary cit- small and secondary cities benefit from direction by the cen- ies benefit from direction by the central government tral government but remain competent stewards of financial inflows, better allocating resources in their administrative but remain competent stewards of financial inflows, areas according to local needs and particularities. Potentially better allocating resources in their administrative underscoring this point is the last cut of the mayor’s wedge areas according to local needs and particularities. analysis, which although not depicted here, finds that scope is strongly and significantly positively associated with job Figure 4.9 Size, direction, and significance of mayor’s wedge coefficients from city size regression analysis Gross Domestic Product Autonomy Jobs Scope Primary Primary * * * Secondary Secondary * * Large Large * * * Small Small * Correlation coefficient (% change) Correlation coefficient (% change) * Statistically significant correlation at the 10% level Source: Authors’ own analysis; dataset used is authors combining Oxford Economics dataset and various indices 43 How does inequality relate to city where αc and βt are city and year fixed effects, respectively, competitiveness? and Ineg_migrate is an identity variable taking the value 1 if the city experiences a negative net migration. The dependent Even without significant increases in administrative powers, variable Gini index measures city inequality for a given city competitive cities are good at reducing inequality—especial- (denoted by the subscript c) at a given point in time (denoted ly cities that attract a huge influx of migrants. Descriptive by the subscript t). results across 750 cities spanning 12 years suggest that rapid city-GDP growth at early stages of incomes is associated with The regression results show that cities with negative migra- larger reductions in inequality—with momentum slowing at tion rates tend to be more equal (table 4.4). In other words, the US$20,000 per capita level and then reversing when cities cities that see an influx of migrants tend to perform worse on reach income levels of US$50,000 to US$60,000 per capita. inequality indicators. This result may be a natural by-product of the process of urbanization, as low-income migrants move We performed a nonlinear multivariate regression analysis to growing cities in search of job opportunities. Under such controlling for year and city fixed effects to test the robust- circumstances, we would expect to observe a high inequality ness of the observation that richer and more competitive ratio. As the income level of a city increases, however, the cities tend to see a decrease in inequality:11 inequality level of a city also seems to drop, bringing shared prosperity to city residents. In the dynamic sense and over (4.2) City Gini Indexct=αc+βt+δCity_GDP_PCct+ time, then, economic development reduces or “mops up” a θCity_GDP_PCct2+Ineg_migrate+εct, city’s initial inequality. Figure 4.10 Inequality by income in 750 OE cities Table 4.4 Regression results of city inequality versus city income level City Fitted values 0.8 City inequality level City Gini index City GDP per capita −0.000683*** 0.6 (0.000183) Citi Gini Index City GDP per capita^2 −9.72e-07 (1.53e-06) 0.4 City with negative immigration −0.00142** (0.000637) Constant 0.470*** 0.2 (0.00205) City GDP per capita (US$, thousands) Observations 8,504 R-squared 0.967 Source: Oxford Economics Dataset Year fixed effects Yes City fixed effects Yes Source: authors’ own analysis; dataset used is oxford econ dataset Significance level: ** = 5 percent, *** = 1 percent. 44 5. Attempting to Arrive at a List of the duced the counterintuitive finding that many cities that did World’s Most Competitive Cities not meet the competitiveness criteria—which were defined at the country level—still performed very strongly by global An attempt to arrive at a short list of the most competitive standards. The phenomenon stems from the fact that even cities provided a cautionary lesson about the challenges relatively rapid growth in cities of Europe and Central Asia, inherent in comparing a world of cities even after attempting Latin America and the Caribbean, or the OECD pales in com- to control for national contexts. parison with the growth rates achieved by even below-aver- age performers in the world’s other regions. Even within East We settled on three criteria to operationalize city competi- Asia and Pacific, the disparities across countries were stark. tiveness. To be considered one of the world’s most competi- tive cities, a city had to outperform its country (because even A future analysis might attempt to define city outperfor- in a globalizing world, a city cannot be taken outside of its mance at the regional level and then determine the world’s country context) on private sector job growth, productivity most competitive cities on a region-by-region basis. However, growth, and household disposable income growth. Growth city competitiveness just may remain difficult to quantify, to had to be positive, meaning that a city could not be deemed generalize across regions of the world, and to demarcate with competitive simply by contracting more slowly than its coun- a clear threshold. After all, none of the other popular city try. Growth rates were measured in annualized terms from competitiveness indexes analyzed in section 3 was found to 2000 to 2012.12 have robust global predictive power either. In the end, 130 of 750 cities outperformed their countries on As a result, we decided to quantify city competitiveness on all three aspects of competitiveness (figure 5.1). The analy- a metric-by-metric basis. Throughout this paper and the sis yielded a list that was on its face intuitive and included accompanying framework paper, we have adopted a simple several cities from the World Bank Group Competitive Cities and straightforward approach to comparing the performance Knowledge Base project’s case study: Bucaramanga (Colom- of competitive cities to their peers: we compare the top 10 bia), Coimbatore (India), and Gaziantep (Turkey). percent of performers on any given metric to the remainder, or at times the top quarter. This methodology allows competi- However, given the diversity of their country and regional tiveness to be determined dynamically and depending on the contexts, the most competitive cities did not look particularly indicator in question—thereby accounting for the reality that competitive at the global level. Comparing the performance competitiveness has many different dimensions, and rarely of the 130 most competitive cities as defined with the re- does one city outshine the rest. maining set of cities on economic outcomes of interest pro- Figure 5.1 The 130 cities that outperformed their countries on all three indicators of city competitiveness Not labeled in Europe and Central Asia Dnipropetrovosk, Kharkiv, Krasnodor, Voronezh Not labeled in OECD Budapest, Copenhage, Honolulu Not labeled in East Asia and Pacific Anshan, Dezhou, Jining, Liaocheng, Tianjin, Xianyang, Xianxiang, Yulin Not labeled in Sub-Saharan Africa Not labeled in Latin America Abeokuta, Benin City, Not labeled in South Asia and the Caribbean Enugu, Ibadan, Kaduna, Ahmedabad, Amravati, Queretaro, San Luis Postoi Lagos, Maiduguri, Oyo, Aurangabad, Coimbatore, Zaria Guntur, Hyderabad, Kozhikode, Thiruvananthapuram, Tiruchi- rappalli, Tiruppur, Warangal Source: Oxford Economics Dataset 45 6. Conclusion of their quality and ability to predict successful economic outcomes. Correlation analyses that explore the relationship This technical overview of global quantitative analytics between determinants of competitiveness, such as institu- attempts to build a knowledge base on city economic per- tions, physical infrastructure, social infrastructure, human formance and city competitiveness. Compared with similar capital, and innovation, helped peel the onion further. Initial efforts, this technical exercise stands out in terms of its findings led to an intriguing hypothesis for further testing: geographic coverage (750 of the largest cities across 140 coun- that cities must square away a succession of building blocks tries); its breadth and depth of topics (urban inequality, job of competitiveness—institutions and social and physical in- creation through tradable sectors and FDIs, key factors that frastructure at lower incomes, then innovation capacity—to drive city competitiveness outcomes, and so on); and, impor- build the human capital base required to compete, grow, and tantly, its identification of global and regional trends perti- prosper as a high-income city. More robust techniques such nent to economic development in cities (city development as fixed-effects regressions were then used to understand pathways, mechanisms linking governance and outcomes, associations and generate findings regarding the mechanisms and the like). that drive competitiveness. For instance, one of the import- ant findings is that a Beginning with the basics, the very first stage of the analyt- larger mayor’s wedge, in ics exercise involves benchmarking a global sample of 750 particular the scope of Cities must square away cities by comparing their performance with each other, with the mayor’s administra- a succession of building regional and national averages, and over time. Interesting tive remit, generally has blocks of competitive- global trends and findings start to emerge. First, cities gen- a positive relationship ness — institutions erate a disproportionately large share of new private sector with city competitive- jobs. In fact, job creation in cities is linked to the broader ness outcomes. However, and social and physical urban inequality debate: cities that observe a large influx of this effect is usually infrastructure at lower migrants looking for jobs tend to see their inequality indi- conditional on capacity: incomes, then innovation cators rise sharply initially, but the evidence presented in the local government’s capacity — in order to this paper posits that as cities develop, inequality drops as capacity determines build the human capital migrants are assimilated into the local labor force. Second, whether mayors can ef- base required to compete, the global data support a particular development pathway for fectively harness and use cities: from market towns to production centers to creative their new powers to bring grow and prosper as a and financial services hubs. At the same time, these path- growth and prosperity to high income city. ways are not etched in stone: for instance, the manufactur- their cities. ing sector is slowing as an engine of growth and structural change, especially for low-income African cities. Third, in We find no holy grail of city competitiveness. This analysis terms of absolute volume, after normalizing using city GDP, has left us, and probably others, wanting. We did not set out two-thirds of the most attractive investment destinations are to, nor did we expect to, uncover singular truths about city in Sub-Saharan Africa, South Asia, and East Asia and Pacific competitiveness. That being said, detailed and comprehensive excluding China. And finally, the tradable sector turns out to data sets focusing on particular regions should produce in- be important for cities not just because it creates more and sights specific to the cases at hand. This work points to global better jobs, but also because it creates a multiplier for jobs trends on topics such as the existence of income traps or of created in the non-tradable sectors. In cities with lackluster earlier deindustrialization, but these trends require further tradable sectors, the non-tradable sectors also lose momen- exploration to substantiate. Nonetheless, all of the research tum. underlines the importance of understanding global trends and basing conclusions about them on reliable data. It also The next stage of the technical exercise included an ex- illustrates the value of city-based analyses, both broad based tensive literature review, followed by a survey of existing and customizable, depending on the client and the question and frequently used city indexes along with an assessment at hand. 46 Notes 1 The World Bank’s 2009 World Development Report (World Bank 2009, 8 Marshallian externalities refer to benefits that firms derive simply by 105) cites sources indicating that 50 percent of world GDP is produced on locating near other firms engaged in similar activities—knowledge spill- just 1.5 percent of the world’s land and that 90 percent of world GDP is overs, labor market pooling, and specialized suppliers—as observed by produced on just 16 percent of land; however, the land area estimate for all Alfred Marshall in Principles of Economics (1890). Jacobian externalities, of the world’s cities is unknown. in contrast, refer to the benefits that firms derive from locating near firms Throughout this study, regions are classified according to the standard 2 engaged in unrelated activities—chance encounters, differing perspectives, World Bank practice. pollination across industries, and serendipitous innovation—as observed by Jane Jacobs in The Economy of Cities (1969). 3 For more on city selection and other methodological questions relating to the data set generally, see OE (2014). 9 For a comprehensive review of literature on the drivers of growth in cities in terms of city size, density, and per capita income, see Duranton and Puga 4 These descriptive statistics do not pretend to summarize the average (2014). For a detailed analysis on the drivers of growth in Colombian cities, characteristics of cities in each region; they merely summarize the 750 cit- see Duranton (2015). ies contained in this specific data set. This section is designed to contextual- ize the subsequent analysis of this data set. 10 According to the literature, fiscal decentralization (that is, local govern- ments having more fiscal autonomy) has a mixed effect across countries. 5 OECD cities are representative cities located in developed countries that This result is probably because fiscal decentralization has to be coupled are OECD members. with political and administrative decentralization when being evaluated, 6 For a thorough review of the ambiguities of the literature, see Görg and because a systematic approach that takes into account the design of fiscal Greenaway (2003). decentralization is more relevant than the fiscal dimension alone. See Mar- 7 The model is similar to the one used in Barro (2015). The conditional con- tinez-Vazquez, Lago-Peñas, and Sacchi (2015). vergence rate of 1.4 percent is from a model without fixed effects, whereas 11 A robust test taking the log of city GDP per capita to account for extremely 9.0 percent is from a model with fixed effects (plus additional controls such values was performed, and similar results were obtained. as education, FDI, and provision of other public services). Positive and sig- 12 For some cities, data did not become available until later in the period. In nificant unconditional convergence rates are obtained as well, at an interval such cases, city performance was assessed over the shorter time span. of 1.9 percent (without fixed effects) and 4.5 percent (with fixed effects). See Barro (2015) for comparison for this convergence exercise. 47 REFERENCES Barro, Robert J. 2015. “Convergence and Modernisation.” Economic OE (Oxford Economics). 2014. “Global Cities 2030 Historical Data: Journal 125 (585): 911–42. Methodology Note.” Oxford, U.K. Duranton, Gilles. 2015. Determinants of city growth in Colombia. Samad, Taimur, Nancy Lozano-Gracia, and Alexandra Panman, eds. Mimeographed, Wharton School, University of Pennsylvania. 2012. Colombia Urbanization Review: Amplifying the Gains from the Urban Transition. Washington, DC: World Bank. Duranton, Gilles, and Diego Puga. 2014. “The Growth of Cities.” In Handbook of Economic Growth, vol. 2B, edited by Philippe Aghion Shapiro, Jesse M. 2006. “Smart Cities: Quality of Life, Productivity, and Steven Durlauf, 781–853. Amsterdam: Elsevier. and the Growth Effects of Human Capital.” Review of Economics and Statistics 88 (2): 324–35 Gennaioli, Nicola, Rafael La Porta, Florencio Lopez De Silanes, and Andrei Shleifer. 2014. “Growth in Regions.” Journal of Economic UN DESA (United Nations Department of Economic and Social Growth 19 (3): 259–309. Affairs), Population Division. 2015. World Urbanization Prospects: The 2014 Revision. New York: United Nations. Görg, Holger, and David Greenaway. 2003. “Much Ado about Noth- ing? Do Domestic Firms Really Benefit from FDI?” IZA Discus- World Bank. 2009. World Development Report 2009: Reshaping Eco- sion Paper 944, Institute for the Study of Labor, Bonn, Germany. nomic Geography. Washington, DC: World Bank. Jacobs, Jane. 1969. The Economy of Cities. New York: Random House. ———. 2013. Planning, Connecting, and Financing Cities—Now: Priorities for City Leaders. Washington, DC: World Bank. Marshall, Alfred. 1890. Principles of Economics. London: Macmillan. World Bank Group. 2015. Competitive Cities for Jobs and Growth: Martinez-Vazquez, Jorge, Santiago Lago-Peñas, and Agnese Sacchi. What, Who and How. Washington, DC: World Bank. 2015. “The Impact of Fiscal Decentralization: A Survey.” GEN Working Paper A2015-5, Governance and Economics Research World Bank and DRC (Development Research Center of the State Network, Universidade de Vigo, Vigo, Spain. Council, the People’s Republic of China). 2014. Urban China: Toward Efficient, Inclusive, and Sustainable Urbanization. Washing- Moretti, Enrico. 2004. “Human Capital Externalities in Cities.” ton, DC: World Bank. In Handbook of Regional and Urban Economics, vol. 4, edited by J. Vernon Henderson and Jacques-François Thisse, 2243–91. Zhu, T. Juni, and Megha Mukim. 2015. “Empowering Cities: Good Amsterdam: Elsevier. for Growth? Evidence from China.” Policy Research Working Paper 7193, World Bank Group, Washington, DC. 48 49 Funding for the companion papers and the main report was provided by CIIP Competitive Industries and Innovation Program Financed by in partnership with www.theciip.org Find the companion papers and the main report at www.worldbank.org/competitivecities