83818 Report No: 83818-TR GOOD JOBS IN TURKEY November 2013 Human Development Sector Unit Europe and Central Asia Region WORLD BANK List of Contributors Executive Summary Rebekka Grun and Sinem Çapar Chapter 1 Rebekka Grun Chapter 2 Meltem Aran and Nazlı Aktakke Chapter 3 Victoria Levin, Tolga Cebeci, Levent Yener, and Altan Aldan Chapter 4 Carola Gruen, Bülent Anıl, and Ayşenur Acar GOOD JOBS IN TURKEY 1 Report No: 83818-TR GOOD JOBS IN TURKEY Prepared by: Rebekka Grun, Cristobal Ridao-Cano, Herwig Immervoll, Sinem Çapar, Victoria Levin, Meltem Aran, Carola Gruen, Levent Yener and Tolga Cebeci November 2013 Human Development Sector Unit Europe and Central Asia Region WORLD BANK 2 GOOD JOBS IN TURKEY CURRENCY EQUIVALENTS (Exchange Rate Effective November, 2013) CURRENCY = TL U$ 1.00 = 2.03 TL WEIGHTS AND MEASURES: Metric System ACRONYMS AND ABBREVIATIONS CPI Consumer price index ECD Early childhood development EU European Union FAO Food and Agriculture Organization of the United Nations GDP Gross domestic product HH Household ILO International Labour Organization LFS Labor force survey LFP Labor force participation LP Labor productivity NACE Nomenclature générale des Activités économiques dans les Communautés Européennes NUTS Nomenclature of Territorial Units for Statistics OECD Organisation for Economic Co-operation and Development SBS Structural Business Survey SILC Survey on Income and Living Conditions SME Small and medium enterprise TFP Total factor productivity TL Turkish lira TOKI Housing Development Administration TUIK Turkish Statistical Institute WAP Working-age population Vice President: Philippe H. Le Houerou, ECAVP Country Director: Martin Raiser, ECCU6 Sector Director: Ana Revenga, ECSHD Sector Manager: Roberta Gatti, ECSHD Task Team Leader: Rebekka Grun, ECSHD GOOD JOBS IN TURKEY 3i CONTENTS Acknowledgements vii Executive Summary ix 1. Conceptual Framework 1 1.1 What are “Good Jobs”? 1 1.2 Where do Good Jobs Come From? 3 1.3 Policy Levers 4 1.3.1 Mobility 5 1.3.2 Trade barriers 5 1.3.3 Investment climate 5 1.3.4 Costs of formality 6 1.3.5 Public services where the market fails 6 Chapter 1 References 7 2. Growth, Employment, and Social Cohesion: Post-Crisis Employment Generation for Men, Women, and Youth in Turkey 9 2.1 Introduction 9 2.2 Drivers of Increased Employment Elasticity of Growth in Post-Crisis Turkey 12 2.2.1 Increase in employment in labor-intensive sectors (the composition effect) 12 2.2.2 Increased employment elasticity within sectors 15 2.3 The Beneficiaries of Employment Generation in the Post-Crisis Period 17 2.4 Trend Changes in the Growth of Employment in the Post-Crisis Period 20 2.5 Was There a Structural Change in Female Labor Force Participation in the Post-Crisis Period? 23 2.5.1 Main trends 23 2.5.2 Employment by education and age 26 2.5.3 Other determinants of female labor force participation 27 2.6 Conclusion and Policy Outlook 28 Chapter 2 References 30 Annex 2.1: Data Sources 31 Annex 2.2: Pseudo-Panel Analysis 32 Annex 2.3: Multinomial Logit Model 38 4 ii GOOD JOBS IN TURKEY 3. Job Creation and Productivity 41 3.1 Introduction 41 3.2 Has Recent Structural Change in Turkey Been Growth-Enhancing? 42 3.3 Job Creation and Productivity in the Non-Agricultural Sector in Turkey 49 3.3.1 What kinds of non-agricultural firms create the most jobs in Turkey? 58 3.3.2 What kinds of firms are most productive in Turkey? 64 3.3.3 The linkage between job creation and firm productivity: do more productive non-agricultural firms create the most jobs? 73 3.4 Are Jobs Being Created in Agriculture Increasingly More Productive? 82 3.5 Conclusion and Policy Outlook 88 Chapter 3 References 90 Annex 3.1: Additional Figures and Tables 91 4. Jobs and Living Standards in Turkey 97 4.1 Introduction 97 4.1.1 Methodology 98 4.1.2 Dynamics at the household level 99 4.2 Levels of and Trends in Living Standards 101 4.2.1 The incidence of low income 101 4.2.2 Material deprivation 103 4.2.3 Discussion 104 4.3 Determinants of Living Standards 106 4.4 A More Detailed Look at the Trends in Household Income 108 4.4.1 Sources of household income 109 4.4.2 Decomposing the changes in living standards 111 4.5 Conclusion and Policy Outlook 113 Chapter 4 References 115 Annex 4.1: Background Figures and Tables 116   GOOD JOBS IN TURKEY 5 iii List of Boxes Box 3.1: Accounting for Structural Change in India and Argentina 44 Box 3.2: Analysis Using the 2005-2010 Structural Business Survey (SBS) 49 Box 3.3: Definitions of Job Creation Rate and Productivity 60 Box 3.4: Methodology 74 List of Figures Figure 1.1: Conceptual framework of the benefits provided by jobs 2 Figure 1.2: Dynamics of good jobs 3 Figure 2.1: Labor force participation (activity), employment, and formality rates by gender, 2005-2011 (%) 10 Figure 2.2: Stock and flow of employment by sector (2005, 2009, and 2011) 13 Figure 2.3: Share of exports to EU and Near and Middle East countries between 2005-2012 17 Figure 2.4: Net employment growth by sector and educational attainment, and gender 18 Figure 2.5: Share of employed in the WAP by gender, age, and informal/formal status 19 Figure 2.6: Recovery impacts, by individual type 21 Figure 2.7: Long-term trends in labor force participation, fertility, wages, and vulnerable employment for women in Turkey 24 Figure 2.8: An upper bound estimate of the added worker effect in Turkey 25 Figure 3.1: Within- and between-sector productivity growth by country group, 1990-2005 43 Figure 3.2: Examples of growth-enhancing and growth-reducing structural change, India, and Argentina 44 Figure 3.3: Correlation between sectoral productivity and change in employment shares in Turkey, 1998-2012 46 Figure 3.4: Correlation between non-agricultural sectoral productivity and change in employment shares in Turkey (1998-2012) 48 Figure 3.5: Number of firms by sector, 2005-2010 50 Figure 3.6: Number of 20+ employee firms by firm size and sector, 2005-2010 52 Figure 3.7: Share of different firm size groups among all 20+ firms 53 Figure 3.8: Distribution of employment across firm size percentiles 54 Figure 3.9: Gross firm entry in and exit out of the 20+ employee firm census and net turnover, 2006-2010 56 6 iv GOOD JOBS IN TURKEY Figure 3.10: Job creation rate, 2006-2010 59 Figure 3.11: Job creation rate distribution by sector, 2010 61 Figure 3.12: Job creation rate by firm size, 2006-2010 62 Figure 3.13: Average firm size by age (based on entry into 20+ firm census), for manufacturing and all sectors, 2008-2010 62 Figure 3.14: Job creation rate by region, 2006-2010 63 Figure 3.15: Value added by sector, 2005-2010 64 Figure 3.16: Value added by firm size, 2005-2010 65 Figure 3.17: Average value-added by firm size and sector intensity in Chile, Indonesia, Morocco, and Turkey 66 Figure 3.18: Average value-added by firm size and sector productivity category in Chile, Indonesia, Morocco, and Turkey 67 Figure 3.19: Total employment by firm size and sector intensity in Chile, Indonesia, Morocco, and Turkey 68 Figure 3.20: Total employment by firm size and sector productivity in Chile, Indonesia, Morocco, and Turkey 69 Figure 3.21: Productivity by sector, 2005-2010 70 Figure 3.22: Productivity by firm type, 2005-2010 71 Figure 3.23: Productivity by firm size, 2005-2010 72 Figure 3.24: Productivity by region, 2005-2010 73 Figure 3.25: Sources of productive labor reallocation in Turkey, 2005-2010 77 Figure 3.26: Sources of productive labor reallocation in manufacturing and services in Turkey, 2005-2010 78 Figure 3.27: Productivity growth decomposition in Turkey and selected countries, manufacturing sector 81 Figure 3.28: Output and employment in agriculture, 1998-2012 82 Figure 3.29: Annual change in the number of employed people (%) 83 Figure 3.30: Map of NUTS 1 regions of Turkey 85 Figure 3.31: Net migration rate and share of agriculture 86 Figure 3.32: Regional productivity, employment and migration in agriculture, 2006-2012 88 Figure 4.1 : Share of in-work households, nationally and by region, 2006-2010 100 Figure 4.2 : Population share living in low-income households, 2007-2010 103 Figure 4.3 : Population share affected by material deprivation, 2006-2010 104 GOOD JOBS IN TURKEY v 7 Figure 4.4 : Material deprivation by income decile, 2007 106 Figure 4.5 : Sources of household income, 2007-2010 110 Figure 4.6 : Sources of household income by quintile, 2010 111 Figure 4.7 : The contribution of labor income, non-labor income, and household structure to changes in living standards between 2007 and 2010 112 List of Tables Table 2.1 : Productivity (sectoral GDP/employment) by sector 13 Table 2.2 : Employment elasticity by sector 15 Table 3.1 : Share of small firms in Turkey and selected countries 51 Table 3.2 : Within-sector dispersion of firm size (coefficient of variation) 55 Table 3.3 : Gross firm turnover rates by sector 56 Table 3.4 : Correlation between entry and exit rates across industries 58 Table 3.5 : The relationship between productivity and labor reallocation in Turkey and its sources, 2005-2010 76 Table 3.6 : The relationship between productivity and labor reallocation within sectors and its sources, 2005-2010 78 Table 3.7 : Short-term relationship between productivity growth and job creation 79 Table 3.8 : Productivity growth decomposition for manufacturing, 2005-2009, average of annual figures 80 Table 3.9 : Annual change of agricultural employment (000’s of people), 2005-2012 84 8 vi GOOD JOBS IN TURKEY --- GOOD JOBS IN TURKEY 9 vii ACKNOWLEDGEMENTS This is a joint study by the World Bank (WB) and the Turkish Ministry of Development (MoD). This report was prepared by a team comprising Rebekka Grun v. Jolk (WB), Victoria Levin (WB), Carola Gruen (WB), Ahmet Levent Yener (WB), Altan Aldan (WB), Tolga Cebeci (WB), Gökhan Güder (MoD), Sinem Çapar (MoD), Meltem Aran (Development Analytics), Nazlı Aktakke (Development Analytics), and Bülent Anıl and Ayşenur Acar (BETAM). Cristobal Ridao-Cano (WB) and Herwig Immervoll (OECD) initiated the dialogue to start this work and conceptualized the components and analytical focus of the report and its chapters. We thank the following advisors for cutting-edge inputs and comments during thematic brainstorming sessions; on chapter 3: Mediha Agar (WB), Mary Hallward-Driemeier (WB), Marc Schiffbauer (WB), William Maloney (WB), Marcela Eslava (Universidad de Los Andes, Bogota), and Bob Rijkers (WB); on chapter 4: Pierella Paci (WB), Sergio Oliveri (WB), Carolina Sanchez (WB), Joao Pedro Azevedo (WB), and Diego Angel-Urdinola (WB). Amy Gautam substantially improved the report through her professional editing. Overall guidance was provided by Martin Raiser (Country Director for Turkey, WB), Roberta Gatti (Sector Manager, Human Development Economics, Europe and Central Asia, WB), and Ana Revenga (Director, Human Development, Europe and Central Asia, WB). 10 viii GOOD JOBS IN TURKEY GOOD JOBS IN TURKEY ix 11 Executive Summary Jobs have direct benefits, such as an the relationship between growth and income and livelihood for the job holder. employment in Turkey, with a particular They can also carry indirect benefits, regard to the participation of different such as increased participation in social groups in the labor market, such society and social networks, or a better as women and youth. It then analyzes living standard for an entire household. where jobs are being created and which On a larger scale, jobs that are filled in activities are the most productive for a productive sector enhance growth the Turkish economy, and assesses if for the economy, contributing to a jobs have moved to more productive virtuous cycle that produces more activities in recent years. Finally, and better jobs in the future. This the report proceeds to measure the report, drawing on the 2013 World impact of different types of jobs on Development Report on Jobs, defines the welfare of an entire household and “good jobs for development” as jobs on the household’s relative position in that are rich in these indirect benefits. the overall income distribution. Every The more indirect benefits (also called chapter contains an outlook towards “externalities”) a job has, the better relevant policies that can support the it is for development. Adapting the impact of Turkey’s good jobs. framework to the context of Turkey, the report focuses on the most relevant Growth, Employment, and Social aspects that would make jobs good Cohesion for the development of Turkey: (i) for living standards, formality, and the Prior to the recent financial crisis, the share of the population that enjoys a Turkish economy recorded a 7.2 percent good income; (ii) for productivity, the annual GDP growth rate over the period structural transformation between and 2002-2006. Capital stock and total factor creative destruction within sectors; and productivity were the main contributors (iii) for social cohesion, and particularly to growth, and GDP per capita reached the increased participation of women US$7,651 in 2006. However, due to the and youth in the labor market. impact of the global crisis, the economy shrank by 4.8 percent in 2009. From This report explores the status and 2007-2012, the annual growth rate effects of good jobs in Turkey’s current was 3.3 percent; the main sources of economy. After a brief account of growth in this period were capital stock economic events since the recent and increasing employment, while total global economic crisis, it examines factor productivity was negative. 12 x GOOD JOBS IN TURKEY Turkey recovered very strongly from quality. The growth of employment the crisis and grew by 9.2 percent took place foremost in the services and in 2010 and 8.8 percent in 2011. formal sectors, and the majority of net However, a relatively weak recovery in employment generation affected both the world economy together with high men and women. This group of workers energy prices and stronger domestic was also mainly university educated. demand contributed to an expansion in the current account deficit. Measures A more detailed analysis of labor force to curb the deficit had implications surveys suggests that the changes in for growth; in 2012, the growth rate overall employment, as well as female was 2.2 percent and GDP per capita employment and improved formality, was US$10,527. From 2007-2012, can mainly be linked to three factors: the share of the agriculture sector in (i) absorption in the agricultural sector GDP decreased to 7.9 percent, the of a significant portion of the unskilled share of industry in GDP recovered to female labor force into informal 19.3 percent, and the service sector employment; (ii) temporary growth reached 72.7 percent. In this period, the in the labor intensive residential main growth contributions came from construction sector; and (iii) older private consumption and fixed capital people remaining in the formal labor investment expenditures. market for longer periods. Currently, the improvement in female labor The rapid growth of GDP and force participation, particularly when employment in the post-crisis disregarding returns to the agricultural period, coupled with upskilling and sector, is not yet significantly above its formalization of employment in pre-crisis trend. the labor market and the increased employment elasticity of growth, Job Creation and Productivity present a very favorable picture of the employment situation in Turkey. Good jobs spur economic growth Between 2007 and 2012, average through their higher productivity. annual employment growth was 3.3 Besides providing an income for their percent, reflecting the creation of over holders, jobs with high productivity 4 million new jobs. The unemployment are considered to be “good jobs for rate was 9.2 percent in 2012, and labor development” insofar as they imply a force participation and employment positive externality of higher economic rates increased in the post-crisis period, growth for the whole society. When jobs particularly among women. become more productive, and when labor in a society is reallocated from Most of the job growth has been of good lower-productivity activities to higher- GOOD JOBS IN TURKEY xi 13 productivity activities, the impact on been only slightly in the direction of economic growth is increased, and enhancing growth. jobs become a more powerful driver of progress. Movement from lower- Overall, it appears that there is labor to higher-productivity activities can reallocation from less productive to occur between sectors as well as within more productive non-agricultural them. Structural change, defined activities. Analysis using firm-level data as movement from low-productivity Structural Business Surveys (SBS) of sectors to higher-productivity sectors firms with 20 or more employees over (typically from agriculture to other the 2005-2010 period demonstrates sectors), can be a potent source of that Turkey is dominated by small economic growth. However, within- services firms, whereas large firms sector improvements in productivity contribute the most to both job creation through capital accumulation, and productivity. Examination of exit technological change, and reallocation and entry, regression analysis, and of labor from low-productivity to high- decompositions demonstrates that productivity farms or firms can be just there is, in fact, labor reallocation as powerful. in Turkey from lower- to higher- productivity non-agricultural activities. Recent structural change in Turkey, This movement of labor is happening particularly the move away from to a certain extent both within agriculture, has been growth- manufacturing and services sectors, enhancing. The most significant change as well as between non-agricultural during the 1998-2011 period was the sectors. flow of labor from agriculture to other sectors; indeed, the employment While agriculture as a whole has share of that sector fell from 41.5 in the lowest labor productivity of 1998 to 25.5 percent in 2011, or by all economic sectors in Turkey, it 38.6 percent. The labor productivity of is important to identify good jobs agriculture in 2011 is still the lowest of within this sector given its continued all sectors, at slightly above a third of importance in terms of output and overall productivity. With this in mind, employment. Even though growth- it is not surprising that labor flows from enhancing structural change can help agriculture to other sectors in Turkey raise Turkey’s overall productivity have been growth-enhancing. However, through movement of labor from apart from the general movement away agriculture to other sectors, agriculture from agriculture between 1998 and will remain an important sector for 2011, movement of labor between years to come. As of 2012, agriculture non-agricultural sectors in Turkey has contributed more than 9 percent of 14 xii GOOD JOBS IN TURKEY GDP and employed close to 25 percent for in-work households, especially in of all workers, and this sector has urban areas and Western provinces. experienced a resurgence since 2007. Among households without regular Region-specific data begin to establish labor market attachment, regional a linkage between productivity and differences were larger and increased job creation. The main finding is that over time. regions with higher agricultural GDP per worker appear to have increased Material deprivation is still widespread their shares of agricultural employment among Turkish households, but and attracted migration, indicating significant improvements have been some agricultural labor reallocation achieved since 2006. At the national towards more productive regions. level, the share of people without access to basic goods declined from 29 Jobs and Living Standards percent in 2006 to 21 percent in 2010. Improvements were uniform across The total number of households with the population; material deprivation at least one member of working age was lower in 2010 than in 2006 for both rose by 10 percent between 2006 and household types and in all regions. 2010; many of these new households Progress has been particularly strong were successfully absorbed into the among rural households in Eastern labor market as the share of in-work provinces, where the share of the households increased as well. The materially deprived population increases in in-work households were declined from 69 percent in 2006 to biggest in the Eastern provinces, where 45 percent in 2010. traditionally fewer households are attached to the labor market. The share of in-work households grew more The global financial crisis had only moderately in the Western provinces, limited impacts on trends in living resulting in smaller regional gaps. standards. The main transmission mechanism of the crisis was through The population share with less than the the labor market. Higher unemployment minimum level of welfare increased rates in 2009, especially in urban areas, slightly. The share of the population contributed to the larger number of living in low-income households households with no regular labor increased slightly from 10.2 percent market attachment. To support people in 2007 to 11.7 percent in 2010. out of work, unemployment benefits However, national averages mask large have increased in 2009, which helped differences across regions and by vulnerable households better cope with household type. The share was lower the crisis. GOOD JOBS IN TURKEY 15 xiii The drivers of low income and material policies that support the observed deprivation in Turkey were similar. favorable labor market developments, Age, education level, and years in particular a better integration of of experience of the breadwinner youth and women into formal work. The mattered. Higher living standards were post-crisis policy package included a 5 also associated with different types percentage points reduction in social of employment, but mainly for non- security contributions for all employees, agricultural jobs. The urban population additional reductions in social security enjoyed higher incomes, but was contributions for youth and female more affected by material deprivation. employees as well as expanded active Regional differences in prices for labor market measures. The evidence in consumer goods, differential access to this report lends cautious support to the housing allowances, and agricultural formalization policy package lowering household production in rural areas are the social contributions for youth. likely to have contributed to this result. The 10th Development Plan continues Labor income was the biggest to prioritize the activation of women contributor to total household income and youth. The objective is to achieve and growth in labor income contributed a labor force participation of women to higher living standards among low- of 34.9 percent by 2018. Planned income households. Higher minimum labor market policies move from wages and crisis-related policy protecting jobs to protecting workers, interventions (for example, reduced which empirically favors easier formal hourly wages to keep people in jobs) job entry for women and youth. This are likely to have contributed to this includes individual account-based trend. Changes in the composition severance payments, increased of households also had a positive coverage of unemployment benefits, impact on living standards. In particular, and spreading temporary work the share of workers contributed contracts. To improve work-life balance, consistently to higher incomes. Changes alternative models such as flexicurity, in non-labor income had a large extended parental leave options, and negative impact on living standards, improved access to child care services offsetting the income-increasing trends will be implemented. of other factors. To increase female labor force Policy Outlook participation, a target of 70 percent gross enrollment of four- to five-year- The Government of Turkey has olds in early childhood education has embarked or is reflecting on several been set. This is a promising measure, 16 xiv GOOD JOBS IN TURKEY as according to the evidence presented learning and skills upgrading in this report, the duty to care for young possibilities. children or elderly family members at home is an important barrier to work Other measures in the 10th for women. Development Plan aim at a broader activation of the workforce. Social • Building on this objective, an benefits are planned to be linked to the expansion of child-care provision, employment agency İŞKUR’s activation especially in urban centers, can be programs and the overall placement helped by a variety of tools. One rate of İŞKUR is intended to increase. In approach tested internationally particular, activation policies together includes training unemployed with lifelong learning programs are women with some previous expected to increase the employability skills and experience to set up of the lower-skilled workforce. their own child-care business. In addition, demand-side or supply- Focusing on the objective of increasing side subsidies, as envisaged by productive jobs, several policies can the Turkish government, can ensure accelerate the movement of labor viability. A job in childcare is also towards more productive activities. viable formal employment for The following have already been motivated women of age 50+, as adopted or are being considered by the they typically have raised their own Government of Turkey: children already. • The initiatives encouraging • An expansion of services for the care employment of women and youth of the elderly can be contemplated through reductions in the employer along the same principles. This share of social security taxes can care can be home-based, allowing potentially facilitate the integration of the elderly to stay in their family existing rural-to-urban migrants into residence. While Turkey still has a productive activities and accelerate generous window of opportunity such mobility, and with it agricultural to prepare for the aging of its shedding, in the future. population compared to Western Europe, duties for home-based elder • Implementation of the action plan care are currently preventing many on combating informality appears women from pursuing formal work. to have borne fruit already, as suggested by the significant entry of • Last but not least, many women firms with 20-49 workers in the 2010 would benefit from continuous SBS firm census. GOOD JOBS IN TURKEY xv 17 • Initiatives expanding the scope for In 2012, the Government of Turkey flexible contracting that have been adopted a NES that addresses some of under consideration in the National the most pressing issues of the Turkish Employment Strategy (NES) economy. One central pillar of the NES can facilitate labor mobility and is increasing educational outcomes and reallocation without jeopardizing improving training opportunities. As workers’ security. shown in this report and in the related literature, there is a strong link between • The reform of severance pay, education and the earnings potential which is very high by international of workers. In particular, adult training standards and might be reducing programs will allow current workers to productive labor reallocation, is one improve their set of skills, giving them of the policies envisioned in the 10th access to higher productivity jobs with Development Plan. better pay. • In agriculture, the availability of a As a second pillar of the NES, public support scheme appears to employment opportunities of have had a supportive effect on underprivileged groups such as youth, formalization and reallocation of women, or the long-term unemployed labor to more productive regions. should be promoted; this in turn is likely to reduce the share of low-income Productivity and formality of a job go households. Given the low earnings hand-in-hand. The government can potential of some of these groups, possibly lower the cost of formality a comprehensive activation of cash by ensuring or strengthening the local transfers could help raise the human presence of the services involved capital and living standards of the in tax and social security benefit entire household. For households with administration. no regular labor market attachment, social transfers have been shown Besides developing an environment to play an important role in raising that allows firms to grow and living standards. Improved targeting create more and better jobs, policy and social transfers that also take into makers could build on the post-crisis consideration regional differences may achievements and help reduce the help to further increase the well-being share of low-income households. of out-of-work households. 18 GOOD JOBS IN TURKEY GOOD JOBS IN TURKEY 1 1. Conceptual Framework 1.1 What are “Good Jobs”? The more indirect benefits, or positive externalities, a job generates, the better The World Development Report 2013: it is for development. For example, Jobs presented a new framework a breadwinner’s job can lift an entire family out of poverty. A job can move for thinking about employment from agriculture towards the expansion (World Bank 2012). While economists of productive off-farm employment, have traditionally focused on how such as manufacturing in cities or development generates jobs, the tourism in the countryside. A job can World Development Report offered a move from a less productive sector and new perspective, asking instead what help expand a more productive one. jobs can contribute to development. A job can provide social insurance to The central argument is that jobs can its holder and also to the jobholder’s have a multitude of direct and indirect household. Finally, a job held by a benefits. woman can lead to higher spending on children’s health and education, on Jobs provide obvious benefits to the average, than a job held by a man. people who hold them, but they can also provide benefits to the people The benefits of jobs can be who interact with the jobholder, such conceptualized in three broad as non-working household members. categories (see Figure 1.1). Jobs can: And jobs can provide indirect benefits (i) provide income and livelihoods, beyond the immediate circles of the and thereby ensure a minimum living worker to the wider society as well. The standard; (ii) increase productivity; and (iii) foster inclusion in society. indirect benefits can be economic, for example welfare and social mobility In the context of Turkey, important of the household, or societal, for results in each of the categories example through social peace and civic include: (i) for living standards, engagement. These are indirect benefits increased formality and a greater share of a job, “spillovers” into broader of the population that enjoys a good networks and society. While direct income; (ii) for productivity, a structural benefits are usually measurable with transformation between and within ease, indirect benefits are not; they are sectors; and (iii) for social cohesion, the “positive externalities” in economists’ enhanced participation of women and speak. youth in the labor market. 2 GOOD JOBS IN TURKEY FIGURE 1.1 Conceptual framework of the benefits provided by jobs Source: Adaptation of framework in World Bank 2012. Most people in the world earn their countries. In this way, productivity main income through a job. An income improvements can be registered enables consumption; more income through movements from agriculture and consumption enable one to enjoy to industry and services (including a higher standard of living. physical migration and urbanization). They can also be registered within Jobs that become more productive sectors if more productive firms hire generate economic growth; job more people than less productive turnover (with less productive jobs firms. disappearing in favor of more productive ones) also generates Finally, jobs enhance participation in economic growth. Job turnover is society. Holding a job confers a certain especially important in developing respect in most societies, such that countries where the dispersion of people feel enabled and entitled when productivity across different sectors holding an occupational title. The is much higher than in industrialized distribution of jobs, especially formal GOOD JOBS IN TURKEY 3 jobs, across regions, social classes, 1.2 Where do Good Jobs and gender is directly correlated with Come From? economic and civil equity across the same dimensions. The integration of Good jobs emerge from a systemic women and youth into the labor market interaction of various economic helps their participation in wider society. processes. Single causal chains are Working women have more influence hard to identify, and a good job can over the allocation of the household result from several drivers, and can budget, and working youth are more in turn drive other processes. The confident and forward-looking. following cybernetic1 graph attempts FIGURE 1.2 Dynamics of good jobs Source: Authors’ elaborations 1- Cybernetics is a transdisciplinary approach for exploring regulatory systems, their structures, constraints, and possibilities. Cybernetics is relevant to the study of systems, including social systems. Cybernetics is applicable when a system being analyzed is involved in a closed signaling loop; that is, where action by the system generates some change in its environment and that change is reflected in that system in some manner (feedback) that triggers a system change, originally referred to as a “circular causal” relationship (Wikipedia). 4 GOOD JOBS IN TURKEY an overview. of different producers and clients and can create good jobs. A country At the center of the dynamics is a with borders open to trade, among circular process (stronger arrows) neighboring countries that are also starting with innovation. This is the open, will embark on a path towards so-called “motor” of the system. More greater specialization and produce innovation leads to the foundation more in its sectors of comparative of start-ups, which can create new advantage. Jobs will move from sectors productive and formal jobs. A presence with a disadvantage into sectors with of productive jobs in a city leads to more a comparative advantage. Firms in “urban pull” and migration towards the the latter sectors will expand. At the city (urbanization). If a city grows too same time, firms in the sectors of much and becomes congested, it can disadvantage will die. This turnover of hamper innovation. A mid-sized city firms, not painless, but usually resulting that grows still increases proximity and in higher average productivity, is called agglomeration: producers can talk to “creative destruction.” users, different producers along a value chain can talk to each other. This is The expanding sectors will now sell to helpful for innovation. global clients, and the people working in these sectors will therefore interact To pick out one example, rural-urban with global clients. This interaction will migration of those in search of better find and bring back new knowledge, for economic opportunities can create example, better production processes. better jobs. Rural-urban migrants The new jobs in the growing sectors typically switch jobs and move out of that have a comparative advantage and agriculture into light manufacturing. To are more exposed to trade are better the extent that the urban sectors are than the jobs given up to move there. more productive, the new jobs in the In the case of Turkey, this means the cities are better than the ones left in expansion of some sectors and formal agriculture. However, it is also possible firms, and, on average, a move away that this transition is not quite painless from agriculture. if the foregone job in agriculture is replaced by a precarious urban job or 1.3 Policy Levers no job at all. Government can influence these Many factors outside the labor market dynamics through a series of policy affect job dynamics. For example, trade levers, for which we provide a few also increases proximity and exchange examples. GOOD JOBS IN TURKEY 5 1.3.1 Mobility often applied to maintain domestic production of certain agricultural crops The advantages of proximity, an or to maintain sectors that would not important side effect of urban be competitive internationally and agglomeration, can best express would disappear with trade. The effect themselves if people and firms are of these subsidies is the same as that of mobile and can physically move where an import tariff, keeping uncompetitive they can be most productive. Mobility sectors alive, and thereby obstructing is a concept that encompasses several the turnover of jobs towards better policy levers. Geographic mobility of jobs. people requires a modern transport network; geographic mobility of firms 1.3.3 Investment climate requires a transparent land planning system that balances environmental When firms are free to enter a market and economic concerns. The without excessive entry barriers in economic proximity of people and the form of licensing costs or difficult firms also requires information and planning permissions, for example, and communication highways. when they are free to grow with equal access to credit and compete based on 1.3.2 Trade barriers genuine customer demand rather than nepotism or guaranteed monopolies, Trade barriers such as tariffs and then growth and investment will product subsidies can restrict usually be where the jobs are best, i.e., trade flows. Removing unnecessary most productive. obstacles to trade, such as import tariffs and quotas, will help to It is also important that firms can integrate the domestic economy compete on equal terms for their labor with the global economy and thereby force, without the public sector offering to realize productivity gains through a more comfortable combination of risks specialization in sectors with a and returns (for example, higher wages comparative advantage. It will also help and no possibility of being dismissed). create jobs that are in touch with the Hiring and dismissal dispositions in the global economy and the global frontier law can influence the turnover of labor of knowledge. and, if too binding, keep people from moving easily into more productive A related example is subsidies for jobs. Dispositions on the legal wage sectors with a comparative international and non-wage costs can create a hurdle disadvantage. Subsidies, in the form to formality that excludes those people of price guarantees, for example, are whose productivity could not more 6 GOOD JOBS IN TURKEY than compensate the minimum wage example, asymmetric information + non-wage package. regarding the quality of care often leads to an under provision of these services. 1.3.4 Costs of formality For example, it is often difficult to distinguish between high-quality and Lowering the mandatory social low-quality child care centers and contributions or the mandatory parents may not be willing to pay the minimum wage of some jobs can make higher fees demanded by high-quality it worthwhile to formalize them. Also, providers. High quality centers will exit making the process easier to register the market and the average quality a job or to register for social insurance provided will fall – despite the fact can trigger the formalization of jobs. For that parents demand high-quality child example, merging the administrative care services. Also, childcare services windows for different social insurance carry a large ‘externality’, i.e. they have benefits into a one-stop shop can positive secondary effects for non- ease this burden on employees. users, for example through the child Finally, outreach to vulnerable and less being well-behaved and a good citizen informed groups, such as informal later on. Providers cannot charge for SMEs or employers in remote regions, these positive secondary effects but can help further convert informal jobs incur the costs to provide them. The to formal jobs. government can support the expansion of child care and elder care services For all legal dispositions, proper in a market-friendly manner through enforcement is crucial. This is easier if demand-side vouchers or conditional the dispositions are simple, transparent, cash benefits. If public demand-side and do not leave room for discretion. subsidies build a viable demand from Discretionary application of rules will beneficiaries, private providers can create an uneven playing field and work supply these services in sufficient against the expansion of good jobs. quality and quantity, and create jobs for the self-employed, entrepreneurs, 1.3.5 Public services where the or employees in larger practices. The market fails expansion of these services is also a good policy tool for the development Expanding services and creating jobs of regions, which may be rich in natural in areas where the market fails is a beauty but lacking in private sector job legitimate motivation for the allocation opportunities. of public funds. Examples of services typically underprovided by the market This report explores the dynamics include childcare and eldercare. For of good jobs and some policy levers GOOD JOBS IN TURKEY 7 to support them. While we briefly Consistent with this focus, Chapter 2 discussed in this chapter five levers covers growth, employment, and social relate to the rest of this report, only cohesion, while Chapter 3 explores covers policies that relate to the employment and productivity, costs of formality and public services especially with regard to sector where the market fails, the two levers transitions. Chapter 4 examines that directly influence productive and the dynamics of employment and formal jobs. The other policy levers will living standards along the income be discussed in the future reports. distribution. Chapter 1 References World Bank. 2012. World Development Report 2013: Jobs. Washington, DC: World Bank. 8 GOOD JOBS IN TURKEY GOOD JOBS IN TURKEY 9 2. Growth, Employment, and Social Cohesion: Post-Crisis Employment Generation for Men, Women, and Youth in Turkey Abstract: The rapid growth of GDP 2.1 Introduction and employment in the post-crisis period, coupled with upskilling and In the period after the 2008 crisis, formalization of employment in Turkey experienced a measurable the labor market and the increased increase in both employment and labor employment elasticity of growth2, force participation. The employment present a very favorable picture of rate increased from 44.9 percent of the employment situation in Turkey. the working-age population (WAP) in However, a more detailed analysis of 2005 to 48.6 percent in 2011, and the labor force surveys suggests that there labor force participation rate increased is not yet reason to assume that these from 50.4 percent to 53.5 percent over changes in the labor market will have the same period. While a higher share lasting effects. From current evidence, of the overall population wanted to the majority of the changes observed work throughout this time period in can be linked to: (i) the agricultural Turkey, job creation grew at a faster sector re-absorbing a significant portion pace. The employment-to-WAP ratio of the unskilled female labor force into (i.e., employment rate) went from 67.4 informal employment; (ii) temporary percent to 69.4 percent for men and growth in the residential construction from 22.8 percent to 28.1 percent for sector; and (iii) older people remaining women. in the formal labor market for longer periods, rather than younger people Women, who traditionally have had entering formal jobs at the beginning low levels of labor force participation of their lives/careers. The improvement in Turkey, particularly saw an increase in female labor force participation, in employment and labor force particularly when we disregard returns participation rates. Female labor to the agricultural sector, is not yet force participation increased from significantly above its pre-crisis trend. 25.8 percent in 2005 to 31.3 percent 2- Defined as the ratio of growth in employment to the growth in GDP. 10 GOOD JOBS IN TURKEY in 2011. The labor force participation 67.4 percent in 2005). It subsequently rates of men, on the other hand, bounced to 69.4 percent after the remained stable between 75.5 and crisis. Female labor force participation 75.9 percent from 2005 to 2011, while and employment rates increased even their employment level dipped to 64.9 through the crisis, partly as a result of percent in the crisis year 2009 (from the added worker effect.3 FIGURE 2.1 Labor force participation (activity), employment, and formality rates by gender, 2005-2011 (%) 80 % % active among men 70 % 60 % % employed among men 50 % % formally employed among men 40 % % active among women 30 % % employed among women 20 % % formally employed among women 10 % 0% 2005 2006 2007 2008 2009 2010 2011 Source: TUIK LFS and authors’ calculations. 3- See Section 2.4 for details on female labor force participation and the added worker effect during the crisis. GOOD JOBS IN TURKEY 11 The quality of jobs improved too, in 0.28 in the 2005-2007 growth periods to terms of formality and skill content. The 0.73 in 2009-2011 to 0.96 in 2010-2012). increases in labor force participation and This means that for the same amount of employment rates for men and women growth generated, more people were went hand-in-hand with increases in the employed. formalization of Turkey’s labor market. Overall, 86.9 percent of net employment Hence, in Turkey in the post-crisis generation between 2005 and 2011 was years, the labor market became in formal employment (i.e., workers more skilled and formalized while with social security coverage), versus concurrently experiencing a higher 13.1 percent in informal employment. employment elasticity of growth. This Over time, there has also been an chapter seeks to understand the drivers “upskilling” of employment in both the of changes in labor force participation, nonagricultural and agricultural sectors: employment, and formalization, about 40 percent of all jobs created as well as the distribution of these between 2005 and 2011 (and about half changes across different groups of the of all jobs in the nonagricultural sector) population, especially women versus were taken up by university graduates. men and youth versus mature people. The next three sections answer three Most of the job creation in the post- main questions. Section 2.2 looks at crisis period was in the private sector.4 The public sector was responsible for changes in employment by sector and 10.4 percent of all new job creation at the drivers behind the increasing and 13.2 percent of formal job creation employment elasticity of growth in between 2009 and 2011. For women, different sectors. Section 2.3 analyzes new employment generation in public net employment growth by types of workplaces was higher: one in five workers in the post-crisis period to formal jobs created for women were in determine who benefited most from the public sector in this time period. employment generation between 2005 and 2011. Finally, Section 2.4 examines Post-crisis growth was job-rich. In the female labor force participation in period after the crisis, the employment Turkey and tries to measure a possible elasticity of growth increased (from structural break.5 4- The Labor Force Surveys (LFS) have information on being mapped to the private or public sector only after 2009; hence, the changes in employment by public/private sector are analyzed only for 2009 and subsequent years. 5- The data sources used are described in Annex 2.1. 12 GOOD JOBS IN TURKEY 2.2 Drivers of Increased made up 16 percent of the increase in employment. In manufacturing, Employment Elasticity of on the other hand, the growth rate Growth in Post-Crisis Turkey of employment during recovery was similar to the stock of employment in Turkey’s growth in the post-crisis this sector (the stock of employment period was job-rich. Real GDP grew by in manufacturing was 20 percent while 9.1 percent in 2010 and 8.7 percent in flow into the sector was 19.6 percent). 2011. This rapid economic growth is partly behind the surge in employment. Typically, agriculture and construction But the more than three-fold increase are the most labor-intensive sectors in in the employment-to-growth elasticity an economy, and significant increases in between 2005 and 2011 suggests that employment in these sectors also bring growth has created relatively more about higher employment elasticity of jobs in the post-crisis period.6 There growth for the overall economy. Since are two drivers of the increase in the we do not have an exact breakdown of employment elasticity of growth: labor and capital costs by sector, which (i) a disproportionate increase in would indicate exact levels of sectoral employment in labor-intensive sectors “labor intensity,” productivity level per (i.e., the composition effect); and (ii) worker in each sector was used as a increased employment elasticity within proxy. Table 2.1 shows the productivity sectors. These are discussed in the next level by sector (GDP/employment). For two subsections. instance, the productivity level was 1.86 in agriculture and 4.06 in construction, 2.2.1 Increase in employment compared to 4.54 in the overall economy in labor-intensive sectors (the in year 2009.7 Low GDP per worker in composition effect) these sectors signals their higher labor intensity. Hence, one reason behind In 2009, 51 percent of workers were the rise in the employment elasticity of employed in services, and this sector growth was likely the disproportionate generated slightly more than a third of increase of employment in the labor- all jobs created between 2009 and 2011. intensive construction and agricultural During the same period, agriculture, sectors, compared to pre-crisis levels. which made up 23 percent of the labor Figure 2.2 provides the stock and flow force in 2009, generated 30 percent of of employment by sector for 2005, 2009, the increase in employment, and the and 2011 (the flow of employment is construction sector, which comprised provided separately for 2005-2011 and only 6 percent of employment in 2009, 2009-2011). 6- For elasticity figures and methodology see Table 2.2 and footnote 14. 7- Productivity calculated using 1998 base GDP numbers in Turkish lira (TL) and employment figures both from year 2009. GOOD JOBS IN TURKEY 13 TABLE 2.1 Productivity (sectoral GDP / employment) by sector Note: The units are expressed in 1998 sectoral GDP (value added) figures in 1,000 TL. Source: Authors’ calculations; GDP numbers from TUIK8; employment levels from TUIK LFS. FIGURE 2.2 Stock and flow of employment by sector (2005, 2009, and 2011) 100 % 80 % 60 % 40 % 20 % 0% 2005 2009 2011 Total Increase in Total Increase in Employment Employment (2005-2011) (2009-2011) Agriculture Manufacturing Construction Services Source: Authors’ calculations from TUIK LFS. 8- GDP in constant prices, taking year 1998 as base year. Source: http://www.tuik.gov.tr/VeriBilgi.do?alt_id=55 14 GOOD JOBS IN TURKEY Increasing global food prices and rising a modest increase in the number agricultural earnings were behind the of construction permits issued after surge in employment in the agricultural the crisis (a rise of 0.4 percent in the sector. The increase in employment in total number of permits), there was the agriculture sector can be explained a considerable increase in the “total by higher earnings in the agricultural area of construction” allowed under labor market (a 21 percent increase in these permits (a 25.2 percent increase hourly real earnings between 2005 and in the post-crisis period compared to 20119) experienced in parallel with the the pre-crisis period11). Of the total surge in global food prices. In the same area of new permits issued in the post- time period, there was a 94 percent crisis period, 98 percent were for increase in the FAO food price index and residential buildings, and 85.5 percent a 63 percent increase in the FAO cereals were funded through private means. price index.10 (Chapter 3 untangles the The government’s share in financing productivity and employment dynamics the construction of newly permitted in the Turkish agriculture sector in more buildings was only 14.5 percent (of total detail.) area of construction) in official statistics – however, anecdotal evidence suggests A residential construction boom in that through the provision of public or Turkey increased employment in the subsidized land and through subsidized construction sector. In the construction credit lines for this sector, the state has sector, growth was mainly fueled by played a large role in supporting the the construction boom in residential construction sector in the post-crisis construction: while there was only years.12 9- Source: TUIK LFS. 10- Source: FAO food price index: http://www.fao.org/worldfoodsituation/wfs-home/foodpricesindex/en/ 11- Source for all construction permit statistics: Central Bank Database, extracting data from TUIK. http://evds. tcmb.gov.tr/cgi-bin/famecgi?cgi=$ozetweb&DIL=TR&ARAVERIGRUP=bie_inyprh1.db In the calculation, the post-crisis period is taken as 2010-2012 and the pre-crisis period is taken as 2005- 2007. It is unfortunately difficult to document public support for the sector with official statistics. Budget data for TOKI 12- or data on public bank credits to the construction sector would help triangulate the observations presented here. GOOD JOBS IN TURKEY 15 2.2.2 Increased employment the construction sector, employment elasticity within sectors elasticity increased from an average of 0.47 in the 2005-2007 period to an A second reason for increased average of 1.16 in the 2009-2011 post- employment elasticity of growth is crisis period.13 Manufacturing also the increased employment elasticity experienced increased employment within each sector. For instance, in elasticity of growth in the post-crisis the agricultural sector, employment elasticity increased from an average of period, although in the services sector -1.7 in the 2005-2007 pre-crisis period there was not a significant change to an average of 2.45 in the period between the pre-and post-crisis 2009-2011 (Table 2.2). Similarly, in periods.14 TABLE 2.2 Employment elasticity by sector Source: Authors’ calculations; GDP numbers from TUIK; employment levels from TUIK LFS. 13- Average elasticities were calculated as simple averages across the years. 14- To calculate employment elasticities by sector, the following formula was used: ∈=(∆o c)/(∆o c) where L represents total employment level and Y is GDP . Therefore growth in employment was divided by growth in GDP . The elasticity then can be interpreted as the percent change of employment for every one percent change in GDP . GDP levels were taken from TUIK’s website. GDP levels were not presented in four main sectors, so the subcategories were aggregated into Agriculture, Manufacturing, Construction, and Services. Lastly, employment levels were calculated using TUIK LFS micro data. For productivity levels, the formula used was: Productivity=Y/L where again Y is GDP and L is total employment. Productivity can be interpreted as the amount of GDP created per worker. 16 GOOD JOBS IN TURKEY The reasons for the increased diplomatic and economic relations employment elasticity of growth in with the Middle East, there have the agriculture sector are discussed been changes in the composition of in depth in Section 4 of Chapter 3. destination countries to which Turkey The increased employment elasticity exports. In pre-crisis years 2005 and in the construction sector could be 2006, exports to the EU were 56 linked to Turkey’s successful export of percent of total exports on average; construction services15 as well as the these dropped to an average of 46 residential construction boom. percent in 2010 and 2011. Meanwhile, the share of exports to Middle East Part of the reason for increased countries increased from 14 percent, employment elasticity of growth in on average, to 21 percent in the same manufacturing may be the shift of time period (see Figure 2.3).16 Cebeci, export markets away from Europe Lederman and Rojas (2013) find that the towards the Middle East, entailing average number of employees is higher a shift in production structure. An increasing amount of labor-intensive for exporters to developing countries products is being produced for more compared to exporters to EU countries, diversified export markets in the post- supporting the point that an increased crisis period. Due to a combination share of exports to the Near and Middle of the reduced demand in European East may be associated with higher export markets and Turkey’s increasing employment. This export has increased in recent years together with a shift to destinations in the Middle East and North 15- Africa, more specifically for the reconstruction of Iraq. Also compare Financial Times: “Turkey Emerges As The True Iraq War Victor,” March12, 2013. Source: TUIK export statistics: www.tuik.gov.tr/VeriBilgi.do?alt_id=12 16- GOOD JOBS IN TURKEY 17 FIGURE 2.3 Share of exports to EU and Near and Middle East countries between 2005-2012 60 50 % share in total exports 40 30 EU Near and Middle East 20 10 0 2005 2006 2007 2008 2009 2010 2011 2012 Source: TUIK export statistics 2.3 The Beneficiaries of graduates. Employment Generation in While growth in the formal services the Post-Crisis Period sector for highly educated workers was the driver in employment growth The employment generated after for both genders, there was a parallel the crisis was of high quality overall. increase in the informal agricultural Increases were mainly in the services sector for women, which explains the sector, in formal employment, and in surge in women’s employment between employment of workers with university 2005 and 2011. Of the net employment degrees. Of the new net employment generated for women, 49 percent generated between 2005 and 2011, 76 was in the formal services sector percent was in nonagricultural sectors and 26 percent was in the informal and close to half (49 percent) came from agricultural sector. The distribution of an increase in the services sector. Half net job creation by gender, educational of all jobs in the nonagricultural sector attainment, formal/informal status, and (48 percent) and 38 percent of all jobs sector of employment is provided in created were taken up by university Figure 2.4.17 17- To calculate the net employment generation, we first calculated the level of employment for 2005 and 2011. Net employment generated was then calculated by subtracting the employment level of 2005 from the employment level of 2011. This was also applied for subcategories of gender and age (i.e., net employment generated for female workers, net employment generated for young workers, etc.). Next we looked at how net employment generated changes for different sectors, cross-tabulating workers by age, gender, and education level. Annex 2.1 summarizes the data sources used in the analysis. 18 GOOD JOBS IN TURKEY FIGURE 2.4 Net employment growth by sector and educational attainment, and gender Employment generation (2005-2011) Employment generation (2005-2011) Female workers Female workers 800000 600000 400000 Nodiploma No diploma 200000 Atleast At leastBasic BasicEducation Education 0 Secondary School Secondary School Agriculture Manufacturing Construction Services Agriculture Manufacturing Construction Services Services Services Agriculture Agriculture Construction Construction Manufacturing Manufacturing -200000 Higher Education Higher Education -400000 Formal Formal Informal Informal Male workers No diploma At least Basic Education Secondary School Agriculture Manufacturing Construction Services Agriculture Manufacturing Construction Services Higher Education Formal Informal Source: Authors’ calculations from TUIK LFS. GOOD JOBS IN TURKEY 19 An important contributor to the be explained by: (i) an increase in the increase in employment, and in effective retirement age (in the formal particular formal employment, was sector); and (ii) entry of older women that mature people stayed longer in into the informal agricultural labor the labor force in this time period. market (informal sector). About 26 percent of net employment growth (and 12 percent of formal There was a systematic formalization employment growth) was attributable for men of all age levels, while for to older workers (50-64 years old). At women there was increased formal the same time, the young (ages 15-29) employment in younger age groups benefited from 13 percent of total net and increased informal employment employment growth (and 29 percent for older cohorts (Figure 2.5). This is of formal job creation), i.e., from a consistent with the finding that women measurable movement from informal in the older age groups re-entered the to formal jobs but little new job creation. informal agricultural sector in this time The overall increase in the growth rate period, tending to increase their overall of employment for the older cohort can rate of informal employment. FIGURE 2.5 Share of employed in the WAP by gender, age, and informal/formal status 100 % 100 % 90 % 90 % 80 % 80 % 70 % 70 % 60 % 60 % 50 % 50 % 40 % 40 % 30 % 30 % 20 % 20 % 10 % 10 % 0% 0% 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 female 2005 female 2011 male 2005 male 2011 Source: Authors’ calculations from TUIK LFS. 20 GOOD JOBS IN TURKEY 2.4 Trend Changes in the Two main models were estimated in this Growth of Employment in the analysis. In Model 1, we only controlled Post-Crisis Period for individual characteristics (yielding 24 cells defined by urban/rural location, To assess the changes in the trend of gender, educational attainment, and age employment growth in the post-crisis group). Model 2 included only workers period compared to the pre-crisis in the labor force and used information period, two sets of pseudo-panel on four sectors and formal/informal regressions of worker categories statusbut omitted location (yielding a were estimated (see Annex 2.1 for the total of 96 cells of workers). Model 1 was methodology and a detailed discussion run with a variety of outcome variables of the results).18 The dependent variable for the individuals in the labor market, in the pseudo-panel regressions is the such as labor force participation, difference between the post-crisis and pre-crisis trend in average growth of employment rates, formal/informal the employment level or rates for each employment, or employment rate by of the cells of worker categories. The sector. In Model 2, on the other hand, growth rates in employment level or there was only one dependent variable rates are computed for each cell in the (the difference between the post-crisis pseudo panel and the average growth and pre-crisis trend in average growth rate of employment for each worker of employment level for each cell). type in the pre-crisis, crisis, and post- While Model 1 is useful for determining crisis periods is calculated.19 Using the post-crisis impact on a variety of this method, the analysis tries to see employment outcomes for different how the growth trend in the post-crisis work groups, Model 2 gives insights period changed compared to the pre- crisis period for different groups of about the sectoral impact of the workers, thus comparing post-crisis post-crisis period. The pseudo-panel recovery impact between these groups regressions were weighted by the size (e.g., men versus women, young versus of each cell in order to extrapolate the middle age). results to Turkey’s WAP .20 The analysis is adapted from “How Did the Great Recession Affect Different Types of Workers? Evidence 18- from 17 Middle-Income Countries” by Newhouse and Cho(2012) in World Development Vol.41 and “Turkey: Managing labor markets through the economic cycle” Report No. 70130 of World Bank ECSHD. The time periods were defined as follows in the quarterly LFS data: Pre-crisis period: Q1 2005 – Q3 2008. 19- Crisis period: Q4 2008 – Q3 2009. Post-crisis period: Q4 2009 – Q4 2011. The pseudo-panel regression was run with 96 observations, and weights were constructed using the size of 20- the worker cell in the baseline quarter (Q1 of 2005) to give more weight to worker types more prevalent in the labor force. Some of the worker categories were too small and for some quarters did not include any workers. Such worker categories were dropped from the regressions. Hence, a total of 85 cells (observations) were used in the pseudo-panel regressions. GOOD JOBS IN TURKEY 21 Several results came out of this pseudo panel analysis, which used quarterly LFS data from 2005 to 2011: FIGURE 2.6 Recovery impacts, by individual type21 Note: Rates given as a share of WAP. Crossed bars are those statistically different from zero. Source: Authors’ calculations based on LFS (TUIK) 21- The full results of Model 1’s pseudo-panel regression are provided in Table A2.2. 22 GOOD JOBS IN TURKEY First, the trend in labor force participation the older age cohort workers compared and employment rate (as a share of the to the middle age cohort workers, while WAP) increased more for workers in in turn no significant difference was urban areas (by 0.8 percentage points found between the impact of recovery and 0.6 percentage points, respectively) on the formal employment rate of the in the post-crisis period, relative to younger age cohort (15-29 years old) the change in trend for workers in and the middle age cohort workers rural areas. In particular, the growth (See Annex 2.2 Table A2.2 Columns 4 rate of employment is higher in the and 5). construction sector, and wage work for this group of urban workers increased The age cohort variables were faster compared to rural workers. important determinants, particularly for formal employment: for those in the The estimates found no statistically older age cohort (50-64 years old), the significant difference in the trends difference between post-crisis and pre- in the employment and labor force crisis growth in the probability of being participation rates of women relative formally employed was 2.4 percentage to men (i.e., the impact of recovery is points higher than the change in trend similar for both groups of workers). for the middle age cohort (the omitted (See Table A2.2 Column 1). One reason category in the regressions), suggesting is that the pre-crisis trends in female a higher impact of recovery for the labor force participation were already older age cohort. However for formal positive. However, for employment employment, there was no significant in the agriculture and construction difference between the impact of sectors, the impact of recovery was recovery for the younger age cohort higher for male workers compared to (15-29 years old) and the impact of female workers. recovery for the middle age cohort. The growth of formal employment The analysis found no significant exceeded that of informal employment difference in the post-crisis trend of in the post-crisis period. The impact formalization for the youngest cohort of recovery in the post-crisis period (ages 15-29) or for women, relative to, was higher for formal employment. respectively, the middle-aged cohort (See Annex 2.2 Table A2.3 Column 1). and men. While in the absolute, the The post-crisis impact of recovery was youngest cohort showed measurable higher on the formal employment rate of movement from informal to formal jobs GOOD JOBS IN TURKEY 23 (28.9 percent of formal job creation population declined from 3.05 children with only 13 percent net job creation), per woman in 1990 to 2.38 in 2000 and their post-crisis trend in growth of 2.09 in 2010. In the same time period, formal employment did not exceed as a result of urbanization and the move the other cohorts (See Annex 2.2 Table away from agriculture, there were lower A2.3 Column 622). For the older age levels of female labor force participation cohort, on the other hand, there was until 2007. At that point, the trend an increase in the growth rate of formal reversed, with increased female labor employment in the post-crisis period force participation through the crisis (a difference of 5.1 percentage points and beyond. in the growth of employment) (See Annex 2.2 Table A2.3 Column 7). This From 1990 to 2008, there was a is consistent with the finding in Section consistent increase in the percentage of 2.3 that older people remain in the labor women (ages 15+) in wage and salaried force for longer periods of time if they work. In fact, through this time period, are formally employed, and this likely the percentage of women in wage contributed to increased formalization and salaried work in Turkey increased in the labor market. consistently (up until 2008), while the percentage of women in vulnerable 2.5 Was There a Structural employment (as unpaid family workers Change in Female Labor Force or self-employed) declined. This trend Participation in the Post-Crisis was stopped by the crisis and the share Period? of women in vulnerable jobs increased. This was likely due to the increased 2.5.1 Main trends female labor force participation in agriculture, especially among older Turkey is undergoing a structural women. Latest numbers suggest transformation in the labor market for that the share of women in wage and women, albeit slowly. It is instructive to salaried work continued to increase take a broader perspective on female (51.6 percent in 2011 and 54.3 percent labor force participation as it relates to in 2012) and that the share of women the percentage of women employed in vulnerable employment decreased in vulnerable work and to fertility rates again (47.2 percent in 2011, 44.4 percent (Figure 2.7). The fertility rate in the in 2012). 22- This result holds when we look at employment rates (as a percentage of WAP) in Model 1, and employment levels in Model 2. The WAP changed by only 2 percentage points between 2005 and 2011. 24 GOOD JOBS IN TURKEY FIGURE 2.7 Long-term trends in labor force participation, fertility, wages, and vulnerable employment for women in Turkey Panel A. Female labor force participation and fertility rate Labor force participation rate percent (Left axis) (Left axis) Panel B. % of women in wage work versus vulnerable employment % Source: World Bank Gender Stats.23 23- See http://databank.worldbank.org/data/views/variableSelection/selectvariables.aspx?source=gender-statistics GOOD JOBS IN TURKEY 25 There is some evidence to suggest of employed married women increased that female labor force participation steadily between 2006 and 2009 and increased during and after the crisis declined thereafter. This can be taken as a result of the added worker as an upper bound on the added worker effect.24 The overall total number of effect since the number includes women married women employed increased who may have already been working through the years, including the crisis years. The proportion of employed before the crisis, and whose husbands married women who had unemployed lost jobs while they did not (see Figure husbands relative to the overall number 2.8). FIGURE 2.8 An upper bound estimate of the added worker effect in Turkey 5,000,000 4.5 % 4,500,000 Husband are unemployed 4.0 % % of employed married women whose Number of employed married women 4,000,000 Husband are employed or out of 3.5 % husbands are unemployed 3,500,000 labor force 3.0 % 3,000,000 2.5 % 2,500,000 2.0 % 2,000,000 1,500,000 1.5 % 1,000,000 1.0 % 500,000 0.5 % 0 0.0 % 2005 2006 2007 2008 2009 2010 2011 Source: TUIK LFS and authors’ calculations. There was no significant structural entered the service sector, a large change in female labor force percentage of women also entered participation in the post-crisis years, the informal agricultural workforce as mainly because the positive trend of unpaid family workers. Total female the pre-crisis years continued. While labor force participation increased more women and university graduates to 31.3 percent in 2011. However, The added worker effect is an increase in labor supply of married women when their husbands become 24- unemployed. 26 GOOD JOBS IN TURKEY without the re-entry of women into the regressions). The post-crisis dummy agricultural labor force (between 2005 variable interacted with the young age and 2011), this level would have been cohort also had a negative correlation 29 percent in 2011 (up from 25.8 percent coefficient for formal employment, at in 2005).25 Chapters 3 and 4 examine - 0.131 (with p-value<0.01), meaning the role of unpaid family workers in that being young was associated with agriculture in more depth, from a a smaller marginal probability of being productivity and welfare perspective, formally employed on average in the respectively. post-crisis period compared to the pre- crisis period (see Annex 2.3 Table A2.4 2.5.2 Employment by education Column 9). and age The middle aged female cohort was In the post-crisis period compared to the more likely to be employed compared pre-crisis period, more highly educated to the young, and their likelihood of women were less likely to be formally being formally employed in the post- employed (compared to women with crisis period also increased in the post- no formal education) while middle aged crisis period compared to the young women were more likely to be formally (see Annex 2.3 Table 2.4 Column 9). employed compared to the young age cohort women. To take a closer look For the older age cohort of women, at the determinants of female labor multiple trends are at work. On average force participation and employment in the sample, women in the age in the post-crisis period in Turkey, a group 50-64 were much less likely multinomial logit regression model to be formally employed compared was estimated. Annex 2.3 provides the to women in the middle age cohort details of this estimation. (the partial correlation coefficient was -2.263, indicating that they were 226.3 Work prospects for young women did percentage points less likely to be not seem to change over the crisis. formally employed compared to the Overall, the youngest cohort (ages 15- middle aged group, controlling for other 29) had a negative correlation coefficient characteristics). This is a function of the when the dependent variable was being early retirement age in Turkey and this formally or informally employed (-0.601 high negative coefficient only applied and -0.284, respectively, in the pooled to the formal sector. For informal sector 25- To get the number of new entrants, first the number of employed women was calculated in agriculture in 2005 and 2011 and then the number for 2005 was subtracted from the number for 2011. Then this number was subtracted from the total number of employed women for 2011. The labor force participation rate was recalculated without taking these new entrants into the total employment of women. This rather simplistic methodology assumed that these women would not have been otherwise employed in the economy, which is not necessarily the case. GOOD JOBS IN TURKEY 27 employment, the partial correlation indicates that for this group of women, coefficient on employment for women having young children was associated in the 50-64 age category was -0.449. with an even more negative correlation with formal employment in the post- 2.5.3 Other determinants of female crisis period than before the crisis. This labor force participation is not surprising given that Turkey made very little progress between 2005-2013 Family composition variables and the in making childcare more affordable number of children and elderly people for working women in urban areas.26 in the household continued to be strong On the other hand, it is interesting determinants of female labor force that after the crisis, urban high-skilled participation. In the pooled sample, women became more likely to enter on average, it was found that having the informal sector with an increased young children (in the age group 0-4) number of young children in the significantly reduced a woman’s labor household (0.302 with p-value <0.01) force participation and employment (see Annex 2.3 Table A2.4 Column 6). probability. For the sample of all The informal sector may be increasingly women, each additional child in the 0-4 serving as a cushion, providing part- age group in the household reduced a time work or otherwise more flexible woman’s probability of employment in work arrangements that the formal the formal sector by 0.466 percentage sector denies even to the high skilled. points (p-value <0.01). The probability is -0.30 (p-value<0.01) for urban high- The number of elderly people in the skilled women, -0.918 (p-value<0.01) household (ages 65+) acted as a for urban low-skilled women, and -0.534 negative correlate of female labor force (p-value<0.01) for rural women. participation and employment, and this seemed to be increasingly the case for For urban high-skilled women, the urban high-skilled women in the post- interaction term between the post- crisis period. For urban high-skilled crisis variable and the number of young women, having elderly people in the children (ages 0-4) had a significant and household was associated with a 7.77 negative value for formal employment percentage point decline in formal (-0.172 with p-value <0.01) (see employment on average in the pooled Annex 2.3 Table A2.4 Column 10). This sample (p-value <0.01). In the post- 26- There have been no public subsidies targeting childcare for the age group 0-4 in this time period. The only ECD program scaled up in this time period was for the kindergarten group (60-month-old children), enrolling children in kindergarten one year before starting primary school. No progress has been observed in public programs that target younger children, through public provision or private provision/public financing. As a result, any expansion in supply of childcare has been through private means and childcare remains unaffordable for the poor and middle classes. 28 GOOD JOBS IN TURKEY crisis period, for urban high-skilled in female labor force participation women, the interaction between the isdue to the re-entry of women into number of elderly people and the post- informal agriculture (usually those in crisis variable was -0.129, indicating the middle age or older cohorts, with that each additional elderly person in no formal training). It was calculated the household reduced the likelihood that the female labor force participation of an urban high-skilled woman being rate would have been 29 percent in formally employed by 12.9 additional 2011 rather than 31.3 percent (up percentage points in the post-crisis from 25.8 percent in 2005) without the period (see Annex 2.3 Table A2.4 development in agriculture. Please refer Column 10). to Chapter 3, section 3.4 and Chapter 4 for a detailed analysis of the importance 2.6 Conclusion and Policy of agricultural employment. Outlook While there has been a gradual change in the structure of female labor force This chapter considered growth and participation over the years, the employment for different social groups evidence does not support a structural in the post-crisis period in Turkey. The break in the post-crisis period compared main findings can be summarized as to the pre-crisis years, mainly because follows. the development of the pre-crisis years was already positive. The changes in the employment elasticity of growth in the post- For youth, a significant change could crisis period can be explained not be found in the employment trend by a disproportionate increase in in the post-crisis period either. For older employment in the agriculture and individuals, an increased probability of construction sectors, which tend to be formal employment was found (likely more labor intensive, and an increase in the result of later retirement). Increased sector-specific employment elasticities. overall formalization in the labor market is partly linked to people staying Most of the job growth has been of good for longer periods of time in formal quality. The growth of employment employment. took place foremost in the services and formal sectors, and the majority of net The Turkish government has already employment generation affected both embarked on several policies that men and women. This group of workers supported the favorable labor market was also mainly university educated. development, in particular a better integration of youth and women into However, a large part of the increase GOOD JOBS IN TURKEY 29 formal work. The post-crisis policy this chapter, the duty to care for young package included a reduction in social children or elderly family members at security contributions for youth and home is important barrier that needs to female employees as well as expanded be overcome by women.27 active labor market measures. The evidence cited in this chapter lends • Building on this objective, an cautious support to the formalization expansion of child-care provision, policy package lowering the social especially in urban centers, can be contributions for youth. The younger helped by a variety of tools. One cohorts have seen a shift from informal approach, first tried in New York to formal work, if with little overall job City (satellite childcare) and later in growth. the United Kingdom (ABC pathway programme) and other countries, The 10th Development Plan prioritizes addresses two problems at once. the activation of women and youth. In this approach, programs train The objective is to achieve a labor force unemployed women with some participation rate of women at 34.9 previous skills and experience to set percentby 2018. Planned labor market up their own child-care business. A policies move from protecting jobs to comprehensive and selective training protecting workers, which empirically course, a start-up credit to upgrade favors easier formal job entry for premises and learning materials, women and youth. This includes and a link with an established individual account-based severance kindergarten (to receive mentoring) payments, increased coverage with are key ingredients for success. unemployment benefits, and spreading Finally, demand-side or supply- temporary work contracts. side subsidies, as envisaged by the Turkish government, can ensure To increase female labor force viability. A job in childcare is viable participation, a target of 70 percent formal employment for motivated gross enrollment of four- to five-year- women age 50+, as they typically olds in early childhood education have raised their own children has been set to facilitate women’s already. employment but also to improve educational outcomes in the long • An expansion of services for the run. This is a promising measure, as elderly can be contemplated along according to the evidence presented in the same principles. While Turkey 27- Further, a profiling of the vulnerable labor force in Turkey shows that the three largest profile groups pertain to inactive women: (i) young urban housewives who never worked; (ii) older urban housewives who never worked; and (iii) young urban mothers in unstable employment. See the companion report World Bank (2013), Activating the Vulnerable into Good Jobs. 30 GOOD JOBS IN TURKEY still has a generous window of productive work or to progress from opportunity to prepare for the less to more productive, especially aging of its population compared formal, jobs. It is important that any to Western Europe, duties for lifelong learning opportunities be home-based eldercare are currently compatible with women’s existing preventing many women from work and family duties. Finally, it is pursuing formal work. Publicly important to build the awareness of subsidized eldercare is still a husbands and other family members relatively underprovided service in about the benefits of life long Turkey and could be expanded and learning for women. staffed with previously unemployed or inactive women and men. These Other measures in the 10th services can be home-based and Development Plan aim at a broader need not require relocation of the activation of the workforce. Social elderly. Again, the expansion can benefits are planned to be linked to be helped through a demand-side İŞKUR’s activation programs and the finance scheme to allow market overall placement rate of İŞKUR is forces to work for the best quality of intended to increase. In particular, provision. activation policies together with lifelong learning programs are expected to • Last but not least, many women increase employability of the lower would benefit from continuous skilled workforce. Also building on learning and skills upgrading education, the 10th Development Plan possibilities. Options to certify foresees increasing tertiary education, existing experience and skills or which could further fuel the jobentry to learn new workplace-relevant observed among the university skills would help women to enter educated. Chapter 2 References Cebeci, T., Y. Lederman and Diego Rojas. Evidence from 17 middle-income 2013. “The Structure of Exports countries.” World Development. across Destinations and Labor-Market World Bank. 2013. “Turkey - Managing Outcomes: An Empirical Case Study of labor markets through the economic Turkey.” Washington, DC: World Bank. cycle.” Washington, DC: World Bank. http://documents.worldbank.org/ Cho, Yoonyoung, and David Newhouse. curated/en/2013/03/17977691/turkey- 2012. “How did the great recession managing-labor-markets -through- affect different types of workers? economic-cycle GOOD JOBS IN TURKEY 31 Annex 2.1: Data Sources For calculation of the employment from the TUIK website (http://www. generation numbers used in the tuik.gov.tr/VeriBilgi.do?alt_id=55). GDP pseudo-panel analysis and multinomial levels are constant prices taking 1998 as logit regressions, we used Turkey’s the base year and total GDP is a sectoral labor force surveys (LFS) from 2005- total calculated by summing the GDP 2011. The LFS is a representative contributions of the agriculture, household survey. The main objective manufacturing, construction, and of the LFS is to obtain information on services sectors. The GDP of each the labor structure of the labor force sector is found by summing the in the country. The survey includes subcategories of each sector. information about economic activity, occupation employment status, For food price indices, the annual food price index of FAO was used (see http:// working hours, earnings, etc. Since it is www.fao.org/worldfoodsituation/wfs- a household survey, it is possible to link home/foodpricesindex/en/). individuals in the household with each other and see, for example, the effect of Construction permit statistics were husbands’ education level on women‘s taken from the Turkish Central Bank’s labor force participation. database, which extracts data from TUIK (see http://evds.tcmb.gov.tr/cgi- For the 2005 survey, 126,704 bin/famecgi?cgi=$ozetweb&DIL=TR&A households answered the RAVERIGRUP=bie_inyprh1.db). questionnaire. This number is 129,527 for 2006; 128,036 for 2007; 129,266 for Lastly, data on fertility rates for Turkey 2008; 135,891 for 2009; 143,871 for were taken from the World Bank’s 2010; and 144,361 for 2011. Gender Stats website (see http:// databank.worldbank.org/data/views/ For productivity and elasticity variableSelection/selectvariables. calculations, GDP levels were collected aspx?source=gender-statistics). 32 GOOD JOBS IN TURKEY Annex 2.2: Pseudo-Panel Analysis Pseudo Panel Regression each cell and the average growth rate of employment for each worker type To assess the changes in the trend of in the pre-crisis, crisis, and post-crisis employment growth in the post-crisis periods was calculated.28 The growth period, two sets of pseudo-panel rates calculated are quarter-on-quarter regressions of worker categories were growth rates (i.e., (Employment level run (see Table A2.1). The dependent in Q4 2011 – Employment level in Q3 variable was the difference between 2011)*100/Employment level in Q3 the post-crisis and pre-crisis trend in 2011).29 Then the difference in growth average growth of e.g., employment post-crisis relative to growth pre-crisis level for each of the cells of worker categories. The growth rate in was calculated and this difference employment level (or labor force was regressed on dummy variables participation rate) was computed for characterizing the different groups.30 TABLE A2.1 Types of workers constructed for the pseudo-panel regression analysis 31 28- Pre-crisis period: Q1 2005 – Q3 2008. Crisis period: Q4 2008 – Q3 2009. Post-crisis period: Q4 2009 – Q4 2011. 29- Year-on-year growth rates would likely yield similar results, because the crisis lasted exactly one year. However, the dataset contains two more quarters in the pre-crisis period than in the post crisis period. 30- In the regression, we also controlled for recovery from the crisis (i.e., the bounce-back effect) by controlling for the rate of growth of the category during the crisis period. 31- In the second model, the cells were not split by urban/rural since very small cell sizes result when interacted with sectors (for instance, there were small cell sizes for urban agriculture and rural construction). GOOD JOBS IN TURKEY 33 Two models were estimated. In Model share of the WAP increased more 1, we only controlled for individual for workers in urban areas (by 0.8 characteristics (yielding 24 cells percentage points and 0.6 percentage defined by urban/rural location, gender, points, respectively) in the post-crisis educational attainment, and age group). period, relative to the change in trend Model 2 also included information on for workers in rural areas. In particular, four sectors and formal/informal status the growth rate of employment is but omitted location (yielding a total of higher in the construction sector, and 96 cells). Model 1 was run with a variety wage work for this group of urban of outcome variables for the individuals workers increased faster compared to in the labor market, such as labor force rural workers. participation, employment rates, formal/ informal employment, or employment After controlling for other factors, rate by sector. In Model 2, on the other the estimates found no statistically hand, there was only one dependent significant difference in the trends in variable (the difference between the employment or labor force participation post-crisis and pre-crisis trend in rates of women relative to men (i.e., average growth of employment level the impact of recovery is similar for for each cell). While Model 1 is useful both groups of workers). However, in determining the post-crisis impact on for employment in the agriculture and a variety of employment outcomes for construction sectors, the impact of different work groups, Model 2 gives recovery was higher for male workers insights about the sectoral impact of the compared to female workers. post-crisis period. The pseudo-panel regressions were weighted by the size The age cohort variables were of each cell in order to extrapolate the important determinants, particularly results to Turkey’s WAP . 32 for formal employment: for those in the older age cohort (50-64), the Model 1 results: changes difference between post-crisis and pre- in employment and labor force crisis growth in the probability of being participation by worker type formally employed was 2.4 percentage points higher than the change in trend In the first set of pseudo-panel for the middle age cohort (the omitted regressions, the trend in labor force category in the regressions), suggesting participation and employment as a a higher impact of recovery for the 32- The pseudo-panel regression was run with 96 observations, and weights were constructed using the size of the worker cell in the baseline quarter (Q1 of 2005) to give more weight to worker types more prevalent in the labor force. Some of the worker categories were too small and for some quarters did not include any workers. Such worker categories were dropped from the regressions. Hence, a total of 85 cells (observations) were used in the pseudo-panel regressions. 34 GOOD JOBS IN TURKEY older age cohort. However for formal By gender: When the sample was employment, there was no significant limited by gender, the impact of difference between the impact of recovery on women’s employment was recovery for the younger age cohort lower in agriculture and construction (15-29) and the impact of recovery for compared to services. The impact of the middle age cohort. Furthermore, recovery for women in the younger for the younger age cohort compared age cohort was also lower compared to the middle age cohort again, the to the middle age cohort women. A differences between post-crisis and pre- statistically significant coefficient was crisis growth trends in the employment not found on the impact of recovery on and labor force participation rates were formal employment for women, which significantly lower, suggesting a lower implies that it was similar to the impact impact of recovery for this group. of recovery on informal employment, while for men the impact of recovery The full results of Model 1’s pseudo- was significantly higher on formal panel regression are provided in Table employment compared to informal A2.2. employment. However, with regard to age categories, a statistically significant Model 2 results: changes in difference between pre-crisis and post- crisis growth trends for men in different employment level by sector and age cohorts could not be found. worker type By education: When the sample was The Model 2 pseudo-panel regression limited to highly educated workers results show that the difference in (high school graduates or higher), it the post-crisis growth relative to pre- was found that this group experienced crisis growth in employment level was a greater impact of recovery on formal significantly higher only for workers sector employment (with the difference in the formal sector when all groups between post-crisis and pre-crisis are included in the regression; the trends 3.1 percentage point above the difference in the growth rate of formal change in trend for informal sector employment in the post-crisis period growth). None of the coefficients were compared to the pre-crisis period was significant for workers with less than 1.5 percentage points above that of the high school education, implying a informal sector. similar impact of recovery for all groups relative to their base groups. The regressions were estimated using various subsets of the data, with the By age cohort: When the sample following results: was limited by age cohorts, it was GOOD JOBS IN TURKEY 35 found that the impact of recovery was Summary of the results of the pseudo- lower for highly educated workers in panel analysis: both younger and older age cohorts compared to low educated workers. • The post-crisis growth in A statistically significant difference employment relative to the pre- between the post-crisis and pre-crisis crisis trend was highest in the formal change in the trend of formalization sector. for the youngest cohort (ages 15-29) compared to the trend in informal • Highly educated workers (high school employment couldn’t be found. (The graduates or higher) experienced same coefficient for the middle aged a stronger impact of recovery on cohort was negative, though not formal sector employment relative significant.) For the older age cohort, to informal employment. on the other hand, a significant positive difference (5.1 percentage points) • For women, the impact of recovery was found in the impact of recovery was lower in the agriculture and on formal employment compared to construction sectors compared to informal employment. This is consistent services and the recovery impact with the finding in Section 2.3 that was also lower for young women older people remain in the labor force compared to middle aged women. for longer periods of time if they are formally employed, and this likely • Formalization gained speed post- contributed to increased formalization crisis for the older age cohort. in the labor market. • The analysis found no significant No evidence was found in this pseudo- difference in the post-crisis trend panel analysis of a significant change of formalization for the youngest in the pre-crisis trend in formal cohort (ages 15-29) or for women, employment for women or youth relative to, respectively, the middle- (age group 15-29) compared to other aged cohort and men. However, groups. in the absolute, the young cohort showed measurable movement from The full results of Model 2’s pseudo- informal to formal jobs (28.9 percent panel regression are provided in Table of formal job creation with only 13 A2.3. percent net job creation). 36 TABLE A2.2 Model 1 Pseudo-panel results (with 24 worker cells) GOOD JOBS IN TURKEY Note: Outcome (dependent) variables are indicated on the left hand side (from 1-18), and independent variables are listed on the right hand side. The average growth rate of the dependent variable in the crisis period is also included in the regressions. The omitted reference categories are: Male, rural, less than secondary school education, and age group 30-49. TABLE A2.3 Model 2 Pseudo-panel analysis results for growth trend of employment by worker type (96 cells) Note: Dependent variable: Y1 = average post-crisis growth of employment for the cell - average pre-crisis growth of employment for the cell. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Average growth of employment in the crisis period is also controlled for in the regression. GOOD JOBS IN TURKEY 37 38 GOOD JOBS IN TURKEY Annex 2.3: Multinomial Logit Model The multinomial logit model used dummy variable was used to indicate allowed for four discrete outcomes whether an observation was from the (i.e., dependent variables) in the 2010 or 2011 datasets. Characteristics labor market: (i) being inactive (the of women, denoted by the vector xi, omitted reference category); (ii) being were: urban location, being married, unemployed; (iii) being informally educational attainment (with no formal employed;33 and (iv) being formally education the omitted category), age employed. Outcomes (ii) to (iv) were group (young or old, with middle age compared to the baseline reference being the omitted category), number category of being inactive.34 of children ages 0-4, number of children ages 5-14, number of adults, The sample data used included four and number of elderly people in the cross-sections of annual LFS data: household. All of these independent the pre-crisis years of 2005 and 2006 variables were then interacted with and the post-crisis years of 2010 and the post-crisis dummy variable to test 2011. The empirical specification was whether there was a change in the as follows: probability of employment in various categories of employment for women Pr(yi=j yi=jor 1)= β0+β1 postcrisisi+β2 with these characteristics in the post- xi+β3 postcrisisi xi+ui crisis period.35 where 1 is the base category (being The interaction between the education inactive); and j indicates the three other variable and the post-crisis variable potential outcomes (being unemployed, shows the change in the probability of being informally employed, or being employment in the post-crisis period for formally employed). A post-crisis women of different education groups. 33- The TUIK definition for formal employment was used: “Are you registered with any social security institution related to this job?” This question is asked to those who say they are working. This is also the definition of job informality used by the government in official statistics. 34- The model is equivalent to a series of pair-wise logit models. 35- The model was first run for different samples of data including urban high-skilled women, urban low-skilled women, and rural women, and also separately for the married women sample where husbands’ characteristics were also included in the regressions (husbands’ employment status and education level). Finally, pre-crisis (2005-2006) and post-crisis (2010-2011) datasets were pooled to test if there was a difference in the labor force participation of women in the post-crisis period. The post-crisis dummy was interacted with all the other explanatory variables to see the effect of post-crisis period. The results of the cross-sectional multinomial logit regressions are available upon request. TABLE A2.4 Multinomial logistic regression, reporting marginal fixed effects Note: Pooled sample of Labor Force Survey Cross-sections for four years of data: 2005-2006 (pre-crisis) and 2010-2011 (post-crisis) for women. Robust standard errors in parentheses. GOOD JOBS IN TURKEY 39 40 GOOD JOBS IN TURKEY .. ~ - - •! -. GOOD JOBS IN TURKEY 41 3. Job Creation and Productivity Abstract: Besides providing an there is some initial evidence on the income for their holders, jobs with regional level of growth-enhancing high productivity are considered to be labor reallocation within this sector. “good jobs for development” in so far as they imply a positive externality of 3.1 Introduction higher economic growth for the whole society. Movement from lower- to Good jobs spur economic growth higher-productivity activities can occur through their higher productivity. between sectors (i.e., through structural As stated in the World Bank’s World change) as well as within them Development Report 2013 on Jobs, (i.e., through capital accumulation, “Economies grow as people get better technological change, skills upgrading at what they do, as they move from farms and reallocation of labor from low- to to firms, and as more productive jobs high-productivity firms). This chapter are created and less productive ones identifies the features of good (i.e., disappear” (World Bank 2012). Thus, highly productive) jobs in Turkey jobs with high productivity, besides and examines the extent to which providing an income for their holders, labor reallocation across and within are considered to be “good jobs for sectors has been in the direction of development” insofar as they imply a more such good jobs. The analysis positive externality of higher economic demonstrates that structural change in growth for the whole society. When jobs Turkey, particularly the movement from become more productive, and when agriculture to non-agriculture sectors labor in a society is reallocated from between 1998 and 2011, has been lower- to higher-productivity activities, growth-enhancing. Moreover, firm- the impact on economic growth is level data reveal that more productive increased, and jobs become a more non-agricultural firms in Turkey appear powerful driver of progress. to be creating more jobs than less productive firms. This movement of Movement from lower- to higher- labor is happening to a certain extent productivity activities can occur within both the manufacturing and between sectors as well as within services sectors, and such reallocation them. Structural change, defined of labor bodes well for the creation as movement from low-productivity of good jobs and future economic sectors to higher-productivity sectors growth in Turkey. Finally, even though (typically from agriculture to other agriculture continues to be the sector sectors), can be a potent source of with the lowest productivity in Turkey, economic growth. This has been 42 GOOD JOBS IN TURKEY demonstrated by the experience of which would imply the reallocation of East Asia and South Asia, where labor from low-productivity to high- movement “from farms to firms” has productivity activities. Finally, the brought with it higher economic growth. chapter looks at the performance of However, within-sector improvements the agricultural sector, which has since in productivity through capital 2007 seen an upsurge in employment, accumulation, technological change, to examine the drivers of productivity in and reallocation of labor from low- to agriculture. It will draw on case studies high-productivity farms or firms can be of regions that have experienced just as powerful. Indeed, McMillan and changes in agricultural productivity Rodrik’s (2011) study shows that labor due to global movements in prices and productivity growth in high-income government policies supporting this countries, Latin America, and Africa sector. in the 1990s and early 2000s can be attributed to within-sector productivity 3.2 Has Recent Structural growth. Change in Turkey Been This chapter identifies the features Growth-Enhancing? of good (i.e., highly productive) jobs in Turkey, and examines the extent to Structural change can be either which labor reallocation across and growth-enhancing or growth-reducing. within sectors has been in the direction In their 2011 paper, “Globalization, of more such good jobs. Since Turkey Structural Change and Productivity still has a significant share of jobs in Growth,” McMillan and Rodrik examine the agricultural sector, the question the relationship between labor flows of structural change remains central and productivity in developing to the country’s development path. countries. The hypothesis is that to Taking this into account, the first stimulate growth and development part of this chapter analyzes whether through structural change, labor should structural change in Turkey from 1998 flow from low-productivity sectors to to 2011 has been growth-enhancing high-productivity sectors. While they or growth-reducing. The second part document this relationship in South focuses on the non-agricultural sector and East Asia between 1990 and 2005, of the economy and uses firm-level the structural changes in high-income data to identify the features of firms countries, Africa, and Latin America that have higher rates of job creation appear to happen in the opposite and productivity, and examines the direction, which implies that structural extent to which jobs are being created change in the latter regions is growth- faster in firms with higher productivity, reducing (Figure 3.1). GOOD JOBS IN TURKEY 43 FIGURE 3.1 Within- and between-sector productivity growth by country group, 1990-2005 -2.00 % -1.00 % 0.00 % 1.00 % 2.00 % 3.00 % 4.00 % 5.00 % Source: McMillan and Rodrik 2011. At the country level, growth-enhancing against average sectoral productivity structural change implies growing (normalized by average productivity employment shares in sectors with in the economy) in the final year. The higher-than-average productivity. resulting bubble graphs, where the size McMillan and Rodrik analyze how of the bubble represents the starting structural change impacts different year’s employment share of the sector, countries by measuring changes in employment shares for major sectors are very revealing (see Box 3.1). 44 GOOD JOBS IN TURKEY Box 3.1 Accounting for Structural Change in India and Argentina In India, agriculture was the sector with the lowest productivity, and the country experienced significant reallocation of labor away from agriculture towards other sectors between 1990 and 2005; therefore, as one can observe by the positive slope of the line in the left panel of Figure 3.2, structural change in India has been growth-enhancing. In general, a picture of growth-enhancing structural change would have more sectors with either i) lower-than-average productivity and falling employment shares (i.e., agriculture (agr) in the case of India), and ii) higher-than-average productivity and rising employment shares (i.e., public utilities (pu), transport and communication (tsc), mining (min), finance, insurance, and real estate (fire), wholesale and retail trade (wrt), construction (con), and manufacturing (man) in the case of India), and less sectors with either i) higher-than-average productivity and falling employment shares (i.e., community, social, personal, and government services (cspsgs) in the case of India), and ii) lower-than-average productivity and rising employment shares. Argentina, in the right panel of Figure 3.2, represents an example where structural change (also away from agriculture towards other sectors) has instead been growth-reducing. This occurred primarily due to even lower productivity of other sectors, such as consumer, government, social, and personal services (cspsgs), and finance, insurance, and real estate (fire) sectors. Indeed, only the agriculture and transport and communication sectors lie in the quadrants of growth-enhancing structural change in Argentina; all other sectors find themselves in the growth-reducing quadrants. Thus, although workers were moving from farms to firms, the average farm appeared to be more productive than the average firm, implying that this movement cannot be expected to bring about higher overall productivity and economic growth.a FIGURE 3.2 Examples of growth-enhancing and growth-reducing structural change, India, and Argentina Correlation Between Sectoral Productivity and Change in Correlation Between Sectoral Productivity and Change in Log of Secrotal Productivity/Total Productivity Log of Secrotal Productivity/Total Productivity Employment Shares in India (1990-2005) Employment Shares in Argentina (1990-2005) Change in Employment Share Change in Employment Share (D Emp. Share) (D Emp. Share) Fitted values Fitted values *Note : Size of circle represents employment share in 1990 *Note : Size of circle represents employment share in 1990 **Note : b denotes coeff of independent variable in regression equation: **Note : b denotes coeff of independent variable in regression equation: In(p/P) - a + bDEmp. Share In(p/P) - a + bDEmp. Share Source : Authots’ calculations with data from Timmer and de Vries (2009) Source : Authots’ calculations with data from Timmer and de Vries (2009) Source: McMillan and Rodrik 2011. a This methodology compares movements between firms with average productivity in each sector; in reality, of course, gains and losses from structural change depend on the marginal productivity of added or lost firms. The methodology also abstracts from changes in population structure (i.e., increases or decreases in working-age population) as well as changes in the labor force participation rates, as it looks only at sectoral shares within the pool of the employed. GOOD JOBS IN TURKEY 45 Recent structural change in Turkey, 1998 and 2012, with the financial and particularly the move away from business services (labeled as “fire” in agriculture, has been growth- the plot) growing the most (by more enhancing. Figure 3.3 replicates than 300 percent, or from 2.5 percent the country-specific analysis in the of all employed in 1998 to 7.6 percent in McMillan and Rodrik paper for the 2012). This sector also had the second- case of Turkey between 1998 (the first highest relative labor productivity, at available year of TUIK data on sectoral 2.7 (after the transport, communication, GDP in 1998 prices) and 2012.36 The and storage (“tsc”) sector, which was most significant change during that at 2.8). “Fire” is an outlier in terms of period was the flow of labor from growth-enhancing labor flows in Turkey, agriculture to other sectors; indeed, and is still a relatively small sector, the employment share of that sector compared to manufacturing (“mnf”), fell by more than 40 percent, from 41.5 retail and wholesale trade (“wrt”), in 1998 to 24.6 percent in 2012. The and community, social, personal, and end-period (i.e., 2012) relative labor government services (“cspsgs”). And productivity of that sector, defined while flows to manufacturing (and as sectoral GDP (in 1998 prices) over transport and communication (“tsc”)) sectoral employment normalized by have been growth-enhancing given total productivity, was the lowest of these sectors’ higher-than-average all sectors at 0.38. With this in mind, it productivity, similarly-sized flows to is not surprising that labor flows from “cspsgs” services (and construction agriculture to other sectors in Turkey (“cstr”)) have been growth-reducing. have been growth-enhancing. With Even larger flows to the trade sector agriculture shedding labor, all other have also been growth-reducing, as sectors with the exception of mining the relative productivity of this sector gained in employment shares between stands at 0.77. 36- This analysis updates the one implemented in Rodrik (2010). 46 GOOD JOBS IN TURKEY FIGURE 3.3 Correlation between sectoral productivity and change in employment shares in Turkey, 1998-2012 Log of sectoral productivity / total productivity in 2012 Change in Employment Share between 1998 and 2012 (share (0.5 = 50%)) Note: Area of circle represents employment share in 1998 Note: Beta denotes coeff. of independent variable in regression equation: In(p/P)=alpha + beta*Change_EmplShare [weighted by 1998 EmplShare] Note: agr - agriculture, utl - public utilities, tsc - transport and communication, min - mining, fire - finance, insurance and real estate, wrt - wholesale and retail trade, cstr - construction, mnf - manufacturing, cspsgs - community, social, personal, and government services. Source: Author’s calculations based on TUIK data and McMillan and Rodrik’s (2011) methodology. The recent reversal of agricultural million (compare Chapter 2). Gürsel shedding in Turkey has not been and Imamoglu (2013) examined the the result of lower productivity in potential reasons for this stark reversal, non-agricultural sectors, but can which preceded and outlived the be attributed to higher agricultural global crisis of 2008-09. The authors prices. The above analysis hides the considered several potential drivers, recent reversal of the movement from including the increase in world food agriculture to other sectors. Agriculture prices, the decrease in non-agricultural began to attract employment in 2007, productivity, and structural changes in such that the share of agricultural agricultural production, such as shifts employment increased from 23.5 towards more labor-intensive crops percent in 2007 to 24.6 percent in (e.g., perennial crops) or production 2012, with the total number of people strategies (e.g., greenhouse usage). employed rising from 4.9 to 6.1 Using regional data on agricultural GOOD JOBS IN TURKEY 47 production and economy-wide relationship between non-agricultural producer prices, Gürsel and Imamoglu labor flows and productivity during the found that the changes in agricultural 1998-2012 period. prices did have a significant positive correlation with changes in regional Only three sectors experienced gains agricultural employment. On the other in non-agricultural employment shares hand, changes in non-agricultural over this time: finance and business incomes (as proxies for changes in labor services, public utilities, and wholesale productivity in non-agriculture), as well and retail trade. The movement of non- as changes in agricultural production agricultural labor from construction had insignificant correlations with and “cspsgs” services to finance and changes in agricultural employment. business services (“fire”) and public utilities (“utl”) has been growth- Apart from the general movement enhancing, but both sectors that away from agriculture between 1998 gained employment are still very and 2012, movement of labor between small (representing 10 percent and non-agricultural sectors in Turkey has 1.2 percent of 2012 non-agricultural been only slightly in the direction of enhancing growth. Figure 3.4 repeats employment, respectively). On the other the Rodrik-McMillan analysis by hand, the movement from relatively excluding agriculture and using only high-productivity non-agricultural non-agricultural employment and sectors (transport and communication, productivity as the denominators. mining, and manufacturing) towards The resulting figure demonstrates a low-productivity wholesale and retail positive but statistically insignificant trade has been growth-reducing. 48 GOOD JOBS IN TURKEY FIGURE 3.4 Correlation between non-agricultural sectoral productivity and change in employment shares in Turkey (1998-2012) Log of sectoral productivity / total non-agric. productivity in 2012 Change in Non-Agric. Employment Share between 1998 and 2012 (of share in 1998) Note: Area of circle represents non-agricultural employment share in 1998 Note: Beta denotes coeff. of independent variable in regression equation: In (p/P_NonAgric)=alpha + beta*Change_NonAgricEmplShare [weighted by 1998 Non-Agric. EmplShare] Note: initials for the denote: min: mining; tsc: transport & communication; mnf: manufacturing; cstr: construction; cspgs: community, social, personal & government services; utl: public utilities; fine: finance, insurance & real estate; wrt: wholesale and retail trade Source: Author’s calculations based on TUIK data and McMillan and Rodrik’s (2011) methodology. While growth-enhancing structural sectors. The next section adopts a change is an important aspect of different approach by utilizing firm- increasing productivity, dynamics level data of non-agricultural firms to within the non-agricultural sector as identify the characteristics of firms with well as within the agricultural sector good (i.e., higher productivity) jobs also need to be examined. The analysis and examining the extent to which above focuses only on one method jobs are being reallocated from lower- of increasing overall productivity – productivity to higher productivity through moving labor across major firms. GOOD JOBS IN TURKEY 49 Box 3.2 Analysis Using the 2005-2010 Structural Business Survey (SBS) This section utilizes the 2005-2010 Structural Business Survey (SBS) dataset compiled by TUIK.The annual SBS dataset is divided into two parts, based on firm employment size: (i) a census of all non- agricultural firms with at least 20 employees (“20+ firms”), and (ii) a sample of non-agricultural firms with less than 20 employees (“20- firms”), with sampling based on region and NACE 4-digit sector. One drawback caused by this sampling methodology is the annual random selection of sampled firms, which makes it impossible to follow a 20- firm over time. In other words, if a firm with less than 20 employees appears in the dataset one year but not in the following year, it is not possible to determine whether that firm shut down or still exists but was simply not included in the sample. Given this drawback of the sample data, the analysis in this study is mainly based on 20+ firms (unless otherwise specified). SBS provides a rich set of information on each firm, including NACE 4-digit sector code, location, employment by gender, investments by function, expenditure by function (including wage expenditure), ownership (foreign versus domestic by share), sales, value added, and material costs. However, no information is provided on the firm’s capital stock or age (i.e., year of registration). While capital stock can be estimated (which is done below to derive TFP), the absence of age information presents a more significant problem, since appearance of the firm in the 20+ part of SBS cannot be identified as either entry or expansion from a smaller-sized firm. Thus, in the analysis below, “entry” is defined as entry into the 20+ employee firm category. Finally, although SBS began collecting data in its current format in 2003, the first two years (2003 and 2004) still had incomplete coverage of 20+ firms, so the analysis starts from 2005. 3.3 Job Creation and then, with services representing more Productivity in the Non- than 80 percent of all non-agricultural firms in the economy (left panel of Agricultural Sector in Turkey Figure 3.5). The dynamics of 20+ employee firms is quite different, with Services is the largest non-agricultural sector in the Turkish economy, but growth falling from 2007 to 2009, and a within the group of 20+ employee significant spike in 2010.37 Moreover, firms, the services sector has outpaced within the 20+ group of firms, services manufacturing only recently. The and manufacturing made up a similar number of firms in Turkey has grown share of total firms until 2010 (right from 2005 to 2007 and decreased since panel of Figure 3.5). 37- According to TUIK staff, the spike in the number of 20+ employee firms cannot be explained by any changes in the methodology of firm listing. Some possible explanations for this spike include formalization of firms due to the government’s efforts to reduce informality, the comprehensive tax amnesty introduced in November 2010. To disentangle real increases in firms versus expansions of existing firms would require their age, which is currently not available (efforts will be made to contact the Ministry of Finance, which holds information on firms’ year of registration). 50 GOOD JOBS IN TURKEY FIGURE 3.5 Number of firms by sector, 2005-2010 Firms of all sizes (census + sample) Firms of 20+ employees (census only) Source: Author’s calculations based on SBS The Turkish economy is dominated hand, within each country, small firms by small firms. Table 3.1 compares are more prevalent in services than Turkey with three developed and three in manufacturing, both as a share of developing countries in terms of the firms and of total employment, with share of small firms (defined as firms France being the only exception for with less than 20 employees) in total the latter. Although this result is valid number of firms and total employment. for Turkey as well, Turkey is an outlier There does not seem to be a relationship for both sectors.38 In fact, small firms between the share of small firms and comprise almost 99 percent of all firms countries’ income level. On the other in services. 38- The apparent contradiction between these results in terms of contribution of small firms to employment and the ones obtained by Ayyagari et al. (2011) can be explained by reliance on different sources (Ayyagari et al. rely on World Bank Enterprise Surveys), different definitions of employment (Ayyagari et al. only counted full- time permanent employees), and different thresholds for small firms (Ayyagari et al. used 5-19 employees, whereas the results above count all firms with less than 20 employees). GOOD JOBS IN TURKEY 51 TABLE 3.1 Share of small firms in Turkey and selected countries Note: Data for Turkey are the average for 2005-2010; for other countries, data are from the 1990s. Source: Bartelsman et al. (2004) and authors’ calculations. Focusing on the 20+ employee firm in mining (56 percent) and the highest census, smaller firms (20-49 employees) in services (68 percent). The largest size comprise more than half of all firms in firms (with 200 or more employees) each sector, and this category grew the represented less than 10 percent of most from 2009 to 2010 (Figure 3.6). In firms in all sectors except mining (where 2010, the lowest share of these firms was they made up 13 percent). 52 GOOD JOBS IN TURKEY FIGURE 3.6 Number of 20+ employee firms by firm size and sector, 2005-2010 Source: Author’s calculations based on SBS 20+ firms are concentrated on the and Morocco. The number of firms in lower side of the size scale in Turkey. this category comprises 62 percent 20-49 employee firms have the of the total number of 20+ firms in largest share not only compared to Turkey. On the other hand, 100-249, other size categories within Turkey 250-499, and 500+ employee firms but also compared to the shares of have the lowest shares in Turkey 20-49 employee firms among 20+ compared to the countries shown in employee firms in Chile, Indonesia, Figure 3.7. GOOD JOBS IN TURKEY 53 FIGURE 3.7 Share of different firm size groups among all 20+ firms % Note: Data coverage is 2005-2010 for Turkey and 1990s and 2000s for Chile, Indonesia, and Morocco. Source: Author’s calculations and Hallward-Driemeier et al. (2013), mimeo Distribution of 20+ employment by and Morocco but lower than that of firm size in Turkey is comparable to Indonesia. The share of the top 1 percent Chile, Indonesia, and Morocco. Over largest firms in total employment of the 2005-2010 period, 55 percent of Turkey was slightly higher than shares total employees worked in the largest of the top 1 percent largest firms in 10 percent of firms in Turkey. This Chile and Morroco but slightly lower rate is almost same as that of Chile than that in Indonesia. 54 GOOD JOBS IN TURKEY FIGURE 3.8 Distribution of employment across firm size percentiles 80 70 60 50 40 % 30 20 10 0 Chile Indonesia Morocco Turkey Top 1% Top 5% Top 10% Source: Hallward-Driemeier et al. (2013), mimeo Within-sector firm size dispersion services, the variation in firm size in Turkey is comparable to other in Turkey is relatively higher than countries. Table 3.2 demonstrates in most comparison countries. In dispersion of firm size by sector some sectors, such as food products (measured by the coefficient of manufacturing, coke and petroleum variation normalized by the cross- products manufacturing, machinery country sector average).39 Coefficients and equipment manufacturing, and greater than 1 indicate that firm size telecommunication and transport dispersion in the corresponding sectors, Turkey’s dispersion of firm country-sector is higher than the size is at least twice the cross-country cross-country average variation in average; on the other hand, publishing that sector. Turkey is between the two and printing and transport equipment extremes of Portugal (lowest) and manufacturing other than motor the U.S. (highest), and is comparable vehicles are much more concentrated to Mexico, Hungary, Romania, and in terms of size than in other countries, France. Both in manufacturing and with coefficients around 0.70. 39- Cross-country sector averages were taken from Bartelsman et al. (2004), which included data from 17 countries (Portugal, France, U.S., Mexico, Romania, Hungary, Slovenia, Korea, Taiwan, Estonia, Brazil, Latvia, Argentina, Italy, Netherlands, Finland, and the U.K.). GOOD JOBS IN TURKEY 55 TABLE 3.2 Within-sector dispersion of firm size (coefficient of variation) (normalized by cross-country sector averages) Note: * Excluding agriculture. Source: Bartelsman et al. (2004) and authors’ calculations. There is very dynamic firm turnover in the 3-8 percent gross firm turnover Turkey. In Figure 3.9, the gross entry rate observed in industrial countries or the is defined as the number of entrant firms 10-13 percent in transition economies, (i.e., those appearing for the first time in as demonstrated in Bartelsman et al. the 20+ employee firm census) divided (2004). (Note however that our analysis by the total number of incumbent can only measure turnover at the 20- and entrant firms in a given year; the and 20+ margin.) exit rate is defined as the number of firms exiting the 20+ employee firm It is also notable that gross firm census in a given year divided by the turnover (entry plus exit rates) is much number of incumbents in the previous larger than net turnover (entry minus year. According to these modified exit rates), and that in some years definitions of “entry” and “exit,” gross (i.e., 2006 and 2010), entry outpaced firm turnover in Turkey was very high exit as in most transition economies, (at 20-40 percent) until 2010, when which could suggest that new firms over 40 percent of firms appeared for are not mere substitutes for old exiting the first time in the 20+ employee firm firms but are entering new, previously census, implying gross firm turnover of nonexistent markets (Bartelsman et al. over 50 percent. This is far and above 2004). 56 GOOD JOBS IN TURKEY FIGURE 3.9 Gross firm entry in and exit out of the 20+ employee firm census and net turnover, 2006-2010 % Source: Author’s calculations based on SBS Firm turnover in Turkey varies (see Table 3.3). In terms of detailed considerably by sector. In general, sectors, gross firm turnover is lowest churning is highest in construction in petroleum products and motor (more than twice the overall rate), vehicles manufacturing, and highest and is about 10 percent higher in in fabricated metal and other transport services than in manufacturing equipment manufacturing. TABLE 3.3 Gross firm turnover rates by sector (entry in and exit out of 20+ employee firm census, weighted by employment, relative to country average) Source: Authors’ calculations. GOOD JOBS IN TURKEY 57 There is evidence that the Turkish size of entering and exiting service economy is characterized firms differs between detailed sectors. predominantly by within-sector Industrialized countries (such as creative destruction. One way Portugal and the U.S. but not France) to assess the main driving force have very high positive weighted behind firm churning is to look at the correlation rates, demonstrating correlation between exit and entry that the sectoral structure in these rates within sectors (Bartelsman et al. countries is settled, and there are very 2004). If there is significant structural few, if any, fast-growing sectors at the change (resources being reallocated expense of other sectors. However, between sectors), then shrinking in transition countries, correlation rates are relatively low (less than 0.2 sectors should have high exit rates in the case of Latvia and Romania) while growing ones should have high and insignificant, implying that some entry rates, leading to a negative sectors must be expanding in these correlation. On the other hand, if countries and some others must be within-sector creative destruction is contracting. With a correlation of 0.28, driving exit and entry, sectors with high Turkey is just above transition countries firm entry should also have high firm but much lower than industrialized exit, and thus a positive correlation countries, suggesting that sectoral will be observed. As shown in Table adjustment might still be ongoing. 3.4 for simple annual correlations, Looking at correlations of period Turkey has significant within-sector averages accounts for the possibility creative destruction in manufacturing that industry changes in entry and exit but not in services. As is the case for do not occur in the same year; this most other countries, adjusting by firm strengthens the evidence for within- size makes a difference for Turkey: the sector creative destruction as the main correlation for services increases and force behind firm exit and entry in non- becomes significant, implying that the agricultural sectors in Turkey. 58 GOOD JOBS IN TURKEY TABLE 3.4 Correlation between entry and exit rates across industries c c Note: * Significant at 10%; a- Entry in/ exit out 20+ employee firm census; b- Turkey: 2005-201, other countries: 1990s. c- Weighted by firm size 3.3.1 What kinds of non-agricultural to be a reallocation of labor from mining firms create the most jobs in and construction towards services and Turkey? manufacturing. Service firms provide the majority The net job creation rate in Turkey in of employment in Turkey. In terms 2010 was about 6 percent. The definition of employment, services make up of net job creation rate is provided in more than 60 percent of total non- Box 3.3. Using this measure, the left agricultural employment, but only half panel of Figure 3.10 plots the overall of non-agricultural employment in 20+ job creation rate in Turkey and the job employee firms. Manufacturing firms creation rate by sector between 2005 and absorb about a third of non-agricultural 2010. Starting at 8 percent in 2006 and employees (and about 40 percent of 2007, the job creation rate declined to employees who work in 20+ employee 4 percent in 2008, then fell dramatically firms). Since 2009, there also appears to negative 5 percent during the 2009 GOOD JOBS IN TURKEY 59 crisis, and rebounded to 6 percent in job creation than entering firms, 2010. The right panel of Figure 3.10 whereas in 2010 the above mentioned decomposes the net job creation of growth in entry of firms with 20 to 49 firms with 20 or more employees into employees contributed more to overall contributions of entering, expanding, job creation than did the expansion of exiting, and contracting firms. One surviving firms. As for job destruction, can observe that until 2010, expanding contracting and exiting firms contribute firms had a more important role in fairly equally to this process. FIGURE 3.10 Job creation rate, 2006-2010 Overall average net job creation rate Average net job creation rate by firm type (census + sample) (census only) 10 40.0 8 30.0 Job creation rate (%) 20.0 Job creation rate (%) 6 10.0 4 0.0 2 -10.0 0 -20.0 2006 2007 2008 2009 2010 -30.0 -2 2006 2007 2008 2009 2010 -4 Contracting firms Exiting firms Entrant firms -6 Expanding firms Overall Source: Author’s calculations based on SBS 60 GOOD JOBS IN TURKEY Box 3.3 Definitions of Job Creation Rate and Productivity In this study, we follow the Davis and Haltiwanger (1999) definition of job creation rate, which is the change in employment between period t-1 and t divided by the simple average of employment in t-1 and t; thus, JCRt = 2 (Lt-Lt-1) / Lt+Lt-1, where JCRt is the net job creation rate (job creation by entering and expanding firms net of job destruction by exiting and contracting firms) at time t, and Lt is employment at time t. This measure has the advantages of being symmetric around zero, lying in the closed interval of [-2, 2], and allowing for estimation of job creation for entering and exiting firms (Davis and Haltiwanger 1999). The analysis below uses two measures of productivity: labor productivity and total factor productivity. Labor productivity is calculated as the firm’s value-added per hour worked. TFP is estimated by the Levinsohn-Petrin procedure within each NACE 2-digit sector, with a firm’s capital stock estimated by using annual investments and the depreciation allowance of firms. Construction firms had the highest Bank and Ministry of Development and service firms the lowest net job 2013, also compare Chapter 2). Finally, creation and destruction rates during in 2010, firms in all sectors created the 2006-2010 period. Construction jobs at the rate of 7 percent, with the firms experienced the highest pre- exception of services where it was 5 crisis job creation rates (or about 15 percent. The frequency distribution percent in 2006 and 2007), followed of the net job creation rate is most by mining and service firms, with dispersed for construction firms, and manufacturing firms creating jobs most concentrated for service firms (as at the rate of 6-7 percent. During the shown in Figure A3.1 in Annex 3.1). This 2009 crisis, manufacturing firms implies that while many construction experienced the most drastic net job firms are creating jobs either through destruction of 8 percent, followed by firm entry or expansion, many are also construction and mining firms at 7 destroying jobs through firm exit and percent; service firms had the least contraction. On the other hand, most net job destruction during the crisis, at of the service sector appears to be 2 percent. This is fairly consistent with creating jobs on net in 2010, with lower the evidence gathered from analysis variance in net job creation between of the Labor Force Surveys (World different firms in the sector. GOOD JOBS IN TURKEY 61 FIGURE 3.11 Job creation rate distribution by sector, 2010 Note: Computation based on 20+ firms only. Source: Author’s calculations based on SBS Large firms have the highest rate of 199 employees, net job creation fell to net job creation in Turkey.40 Firms zero. For smaller firms (i.e., firms with with 50 or more employees created 20 to 49 employees), job destruction obs at the rate of 15 percent or more exceeded job creation in 2008 and in 2006 and 2007, while firms with 20- the gap increased dramatically in 2009 49 employees had a net job creation to negative 10 percent; by 2010, job of only 3 percent (Figure 3.12). During creation in these firms was similar to the 2009 crisis, the gap between job job destruction, which can be attributed creation and job destruction narrowed to many firms still experiencing net job significantly for firms with 200 or destruction, while some firms started more employees (from 11 percent to recovering from the crisis (Figure 3 percent), and for firms with 50 to A3.1). 40- This result is consistent with Ayyagari et al. (2011). 62 GOOD JOBS IN TURKEY FIGURE 3.12 Job creation rate by firm size, 2006-2010 20 15 Job creation rate (%) 10 5 0 2006 2007 2008 2009 2010 -5 -10 -15 20-49 50-100 200- Surviving firms have been only slightly creation from the perspective of post- larger than entering firms until 2010, entry survival and growth rates of firms when firm size at entry decreased in transition economies by comparing substantially. Bartelsman, Haltiwanger, firm size at different ages to firm size and Scarpetta (2004) examine job at entry. As noted above, the SBS FIGURE 3.13 Average firm size by age (based on entry into 20+ firm census), for manufacturing and all sectors, 2008-2010 100 Average firm size 80 60 40 20 0 Manufacturing All sectors Manufacturing All sectors Manufacturing All sectors 2008 2009 2010 age=1 age=3 age=5 Source: Author’s calculations based on SBS GOOD JOBS IN TURKEY 63 dataset does not contain firms’ year of difference is very small, it may be registration, so age can only be defined affected by the economic conditions of from entry into the 20+ employee firm the respective foundation years. Three- census. Still, Figure 3.13 examines year-old firms were established in 2008, how much surviving firms with 20 or which was a strong growth year until more employees grow in size after they the last quarter. appear in the SBS census. By the age of three, the average manufacturing Net job creation rate has been highest firm in 2008 has 61 employees, which in the East region of Turkey and is only four more than the average lowest in the West and Istanbul.41 The entering manufacturing firm in that East region of Turkey has experienced year (the difference for all sectors is the highest rates of net job creation eight employees). By 2009, firm size throughout the 2006-2010 period at entry rises, and so does firm size (with 12-14 percent net job creation in at the age of three years, especially 2006-07); it was the only region during for non-manufacturing firms. Finally, the 2009 crisis not to have net job in 2010, firm size at entry falls, since, as seen earlier, many firms with 20-49 destruction (Figure 3.14). On the other employees enter the market, bringing hand, the West region and Istanbul saw down the average firm size at entry the lowest pre-crisis net job creation and increasing the gap between the rates (at 7 percent), and the most net size of entering and surviving firms. job destruction in 2009 (at negative Interestingly, firm size at age five is 7 percent and negative 6 percent, below that at age three. While the respectively). FIGURE 3.14 Job creation rate by region, 2006-2010 20 15 Job creation rate (%) 10 5 0 -5 -10 East İstanbul Middle North South West 41- The correspondence of provinces to regions is as follows: East (Agri, Erzurum, Gaziantep, Malatya, Mardin, Sanliurfa, Van), Istanbul (Istanbul), Middle (Ankara, Kayseri, Kirikkale, Konya), North (Kastamonu, Samsun, Trabzon, Zonguldak), South (Adana, Antalya, Hatay), and West (Aydin, Balikesir, Bursa, Izmir, Kocaeli, Manisa, Tekirdag). 64 GOOD JOBS IN TURKEY 3.3.2 What kinds of firms are most construction firms was hit hard by productive in Turkey? the 2009 global crisis, and in the case of services, it has not recovered in Service firms have the highest value- 2010, whereas manufacturing and added in the Turkish economy; mining firms’ value-added does not however, within the 20+ employee seem to be affected by the crisis. For firms, the value-added of manufacturing firms with 20 or more employees, the firms is slightly above that of services. picture looks different, with the value- As demonstrated in the left panel of added of services and manufacturing Figure 3.15, services firms contribute firms moving very close together, both the highest value-added to the economy decreasing during the 2009 crisis and of Turkey, although manufacturing recovering in 2010, with manufacturing firms have been catching up. Notably, rebounding more strongly (right panel the value-added of services and of Figure 3.15). FIGURE 3.15 Value added by sector, 2005-2010 Firms of all sizes (census + sample) Firms of 20+ employees (census only) 120 70 Value added (billion of 2003 TL) Value added (billion of 2003 TL) 100 60 50 80 40 60 30 40 20 20 10 0 0 2005 2006 2007 2008 2009 2010 2005 2006 2007 2008 2009 2010 Mining Manufacturing Construction Services Mining Manufacturing Construction Services GOOD JOBS IN TURKEY 65 Almost half of all value-added is which absorb almost half of all non- contributed by firms with 200 or more agricultural employment but produce employees. Despite employing less less than a third of all value-added. than 30 percent of all non-agricultural Notably, while the share of value- workers, firms with 200 or more added contributed by firms with employees contribute almost half 50 or more employees has been of all value-added in the Turkish gradually rising, that of smaller firms economy (Figure 3.16), providing has been falling (with the exception of initial suggestive evidence that firms with 20-49 employees, which their productivity is higher than that saw a rise in value-added share of smaller firms. The opposite is true from 8 percent in 2009 to 12 percent of firms with less than 20 employees, in 2010. FIGURE 3.16 Value added by firm size, 2005-2010 Source: Author’s calculations based on SBS 66 GOOD JOBS IN TURKEY In Turkey, the average value-added resource-intensive sectors have a of firms in capital-intensive sectors higher average value-added. Turkey’s is higher than the value-added of pattern is similar to that of Chile firms in labor-intensive and natural and Indonesia for all size groups resource-intensive sectors for all firm except 250-499 and similar to that size categories except 250-499. For of Morocco for the 250-499 firm size this size category, firms in natural group. FIGURE 3.17 Average value-added by firm size and sector intensity in Chile, Indonesia, Morocco, and Turkey Note: Firm size categories represented by each value in the X axes are as follows: 1 : 20-49, 2: 50-99, 3: 100-249, 4: 250- 499, 5: 500+. Data coverage is 2005-2010 for Turkey and 1990s and 2000s for Chile, Indonesia, and Morocco. Source: Authors’ calculations for Turkey and Hallward-Driemeier et al. (2013), mimeo GOOD JOBS IN TURKEY 67 Firms in high productivity sectors, by size category increases. Although this OECD classification, have the highest ranking is not surprising and is similar to average value-added for all size groups those observed for Chile and Indonesia in Turkey. High productivity sectors are in Figure 3.18, in Turkey, the difference followed by medium productivty and between the value-added of firms in low productivity sectors in average high productivity sectors and medium value-added. The average value-added productivity sectors is smaller than the gap widens in absolute terms as firm differences in Chile and Indonesia. FIGURE 3.18 Average value-added by firm size and sector productivity category in Chile, Indonesia, Morocco, and Turkey Note: Firm size categories represented by each value in the X axes are as follows: 1 : 20-49, 2: 50-99, 3: 100-249, 4: 250- 499, 5: 500+. Data coverage is 2005-2010 for Turkey and 1990s and 2000s for Chile, Indonesia, and Morocco. Source: Authors’ calculations and Hallward-Driemeier et al. (2013), mimeo 68 GOOD JOBS IN TURKEY Quite different from the cases of category is small compared to that of Chile, Indonesia, and Morocco, almost other countries. The total number of equal numbers of employees work in workers in 50-99 and 250-499 firms as capital-, labor-, and natural resource- well as 20-29 and 100-249 firms are intensive sectors, and in all firm very close to each other. In the 500+ size groups in Turkey. For Chile and employee firms, the largest employer Indonesia, more employees work in group, there are only 2.3 times as many natural resource-intensive sectors than workers as in the smallest group with in capital- and labor-intensive sectors 50-99 employee firms. The differential for all size groups. Another interesting between the total number of employed observation from Figure 3.19 is that between largest and smallest size the discrepancy between the number groups is around 5 times for Indonesia of people employed in each firm size and 3 times for Morocco and Chile. FIGURE 3.19 Total employment by firm size and sector intensity in Chile, Indonesia, Morocco, and Turkey Note: Firm size categories represented by each value in the X axes are as follows: 1 : 20-49, 2: 50-99, 3: 100-249, 4: 250- 499, 5: 500+. Data coverage is 2005-2010 for Turkey and 1990s and 2000s for Chile, Indonesia, and Morocco. Source: Authors’ calculations and Hallward-Driemeier et al. (2013), mimeo GOOD JOBS IN TURKEY 69 In Turkey, firms in low productivty highest number of employees except sectors, by OECD classification, employ for in the 500+ size category. For this more people than medium productivity category, low and high productivity sectors and high productivity sectors sectors employ the same number for all size groups except for 500+. In of people. Sector productivity and the 500+ category, equal number of employment share patterns seem to employees work in high productivity be the most favorable in Chile, where and low productivity sectors. The many more people are employed in pattern in Turkey is somewhat similar to high productivity sectors than in other that seen in Morocco. Low productivity productivity groups for firm size groups sectors in Morocco employ by far the larger than 100. FIGURE 3.20 Total employment by firm size and sector productivity in Chile, Indonesia, Morocco, and Turkey Note: Firm size categories represented by each value in the X axes are as follows: 1 : 20-49, 2: 50-99, 3: 100-249, 4: 250- 499, 5: 500+. Data coverage is 2005-2010 for Turkey and 1990s and 2000s for Chile, Indonesia, and Morocco. Source: Authors’ calculations and Hallward-Driemeier et al. (2013), mimeo 70 GOOD JOBS IN TURKEY Service firms have the highest than 10 percent above average in labor productivity in Turkey, while 2008, and then began a steady decline, construction and manufacturing bringing it to 10 percent below average firms have the highest total factor by 2010. Finally, mining had the lowest productivity. Figure 3.21 plots sectoral labor productivity from 2006 to 2009, productivity normalized by the average but in 2010 experienced a significant productivity in Turkey in that year comeback. The dynamics of sectoral (productivity distributions by sector comparisons of TFP appear to be quite for 2010 are reported in Figure A3.2). different, perhaps due to measurement Services firms have higher-than- issues. Only construction firms have average labor productivity, although had consistently above-average TFP , this relative productivity has been with manufacturing firms catching up to declining gradually from 2005 to 2010. them in 2009 and 2010. TFP of services Labor productivity in manufacturing firms started at slightly above average firms was fairly stable at about 10 in 2005, but declined throughout the percent below the average until 2009, period. Mining firms have had the when it rose significantly. By contrast, lowest relative TFP , at 40 to 60 percent labor productivity in mining was rising below average during the period under in the pre-crisis period, peaked at more study. FIGURE 3.21 Productivity by sector, 2005-2010 Labor productivity Total factor productivity GOOD JOBS IN TURKEY 71 Incumbent firms have the highest labor labor productivity of entrant firms had productivity in Turkey, while firms that the highest volatility. Interestingly, lose workers either through contraction TFP comparisons reveal that at the or exit tend to have the highest TFP . beginning and the end of the period The labor productivity of incumbent under study, exiting and contracting (expanding or contracting) firms was firms had the highest relative TFP; above average for virtually all of the expanding firms had the lowest TFP 2006-2010 period. Exiter firms had a between 2007 and 2009. The relative labor productivity that was 5 percent TFP of entrants was even more volatile or less below the average, while the than their labor productivity. FIGURE 3.22 Productivity by firm type, 2005-2010 Labor productivity Total factor productivity Large firms are much more productive the average firm, with some decline in than smaller firms in Turkey. Figure this relative productivity in 2007 and 3.23 presents relative labor and TFP 2008, and recovery in 2009 and 2010. productivity of firms of different sizes. In terms of TFP, the productivity of large In both cases, the productivity of firms firms was 90 percent above average with 200 or more employees is far in 2005 and 2006, fell gradually to 60 above that of smaller firms. In terms percent above average by 2009, and of labor productivity, large firms are then rebounded to 80 percent above about 40 percent more productive than average by 2010. 72 GOOD JOBS IN TURKEY FIGURE 3.23 Productivity by firm size, 2005-2010 Labor productivity Total factor productivity Labor productivity is consistently the lowest labor productivity. The highest in Istanbul, and lowest in the dynamics of regional relative TFP are North region; the regional picture much more complicated, with sharp of TFP varies over time. Labor swings by regions from year to year, productivity of firms in Istanbul is although the frequency distribution about 20 percent higher than the of TFP in 2010 demonstrates that average; firms in the Middle region Istanbul firms have somewhat higher also have higher-than-average labor productivity, and firms in North and productivity in 2009 and 2010 (left East somewhat lower productivity than panel of Figure 3.24). Firms in firms in other regions (right panel of the North and East of Turkey have Figure 3.24). GOOD JOBS IN TURKEY 73 FIGURE 3.24 Productivity by region, 2005-2010 Labor productivity Total factor productivity 3.3.3 The linkage between job Istanbul has the lowest level of job creation and firm productivity: do creation but the highest level of labor productivity. To what extent are the more productive non-agricultural two phenomena – job creation and firms create the most jobs? productivity – linked in Turkey? In other words, are more productive firms Is there a linkage between the firms creating the most jobs, thus facilitating that produce the most jobs and those labor reallocation to these better jobs with higher productivity in Turkey? So for development? The analysis below far, among other findings, firm-level attempts to address these questions data have shown that large firms have by looking at the relationship between the highest net job creation and also job creation and firm productivity using the highest productivity, that services cross-sectional regressions, panel firms have lower net job creation than regressions, and decompositions. other sectors and lower-than-average The methodology for the analysis is productivity in recent years, and that described in Box 3.4. 74 GOOD JOBS IN TURKEY Box 3.4 Methodology Cross-section regressions There is a short-run tradeoff between productivity and job creation. In the medium to long run, more jobs could be created in activities that are more productive and competitive. Since productivity is measured as value-added per worker (or per hour worked), in the very short run, job creation necessarily depresses a firm’s productivity. Thus, one has to look beyond this immediate relationship to a longer time horizon. As the SBS dataset spans the period of 2005 through 2010, the focus of the analysis is the relationship between firm productivity in 2005 (i.e., initial productivity) and its net job creation over the entire period of 2005 and 2010. This implies measuring job creation over six time periods, which can be argued to be a sufficiently long time horizon to abstract away from the short-run tradeoff between productivity and job creation. The focus on firms that survive for the whole 2005-2010 period demands an adjustment for selection, as more productive firms are more likely to survive and appear in our database. As one can observe from the findings in the top panel of Figure 3.8, surviving firms have higher or lower productivity than entering or exiting firms, respectively. Thus, focusing on these firms in our analysis needs to take into account that more productive firms are more likely to survive and self-select into our sample. Not taking this into account in the regression analysis might result in capturing a spurious relationship between job creation and productivity. To correct for this sample selection problem, our analysis follows Comola and de Mello (2009) by using a multinomial logit, where the first stage predicts whether a firm is a survivor, entrant, exiter, or both an entrant and an exiter during the period under study, and the second stage regresses net job creation on initial productivity as well as a set of other firm characteristics, including exporter status (which is correlated with both productivity and net job creation) and four selection terms calculated from the first stage. The first stage uses initial productivity (either LP or TFP), exporter status, and NUTS2 region dummies (the latter as the excluded variables from the second stage) to predict survival. Panel regressions While cross-sectional analysis allows us to look at the overall relationship between productivity and job creation and to assess the extent of productive labor reallocation across sectors, across sub-sectors, and within sub-sectors, we rely on panel regressions to examine how productivity changes within a firm over time affect job creation. To ensure that the results detected on the productivity growth-job creation relation do not depend the initial productivity level of a firm, we control for the productivity level of a firm in the previous year. The sample in these regressions includes all firms that exist for two consecutive years. Decompositions (manufacturing sector) We also decompose manufacturing productivity growth into five components, using the methodology of Foster, Haltiwanger, and Krizan (FHK) (2001) as implemented in Bartelsman et al. (2004) to benchmark the sources of that growth against industrialized and developing countries. FHK decompose productivity growth into five components: 1. The ‘within-firm effect’ is within-firm productivity growth weighted by initial output shares. 2. The ‘between-firm effect’ captures the gains in aggregate productivity coming from the expanding market of high productivity firms, or from low-productivity firms’ shrinking shares weighted by initial shares. 3. The ‘cross effect’ reflects gains in productivity from high-productivity growth firms’ expanding shares or from low-productivity growth firms’ shrinking shares. 4. The ‘entry effect’ is the sum of the differences between each entering firm’s productivity and initial productivity in the industry, weighted by its market share. 5. The ‘exit effect’ is the sum of the differences between each exiting firm’s productivity and initial productivity in the industry, weighted by its market share” (Bartelsman et al. 2004). GOOD JOBS IN TURKEY 75 Cross-section results As for TFP , although the selection term for surviving firms is positive, Overall, productive firms appear to it is insignificant; on the other hand, be creating the most jobs in Turkey. coefficients for other firms are either The results in Table A3.1 suggest that negative (implying that entrant firms firms with higher initial productivity are are less likely to create jobs) or not creating jobs at a higher rate than firms significant. with lower productivity. This holds both for labor productivity (LP) as well as Within-subsector reallocation of labor TFP , although the magnitudes of the appears to be the most productivity- coefficients are quite different, perhaps enhancing in Turkey. Introducing sector due to different distributions of these fixed effects (broad and NACE 2-digit two measures: if labor productivity subsectors) allows us to distinguish increases by 1 percent, the net job between different sources of labor creation rate is expected to increase reallocation and productivity. Three by 14.3 percentage points; if TFP sources of labor reallocation are increases by 1 percent, the net job defined: creation rate is expected to increase by 2.4 percentage points (columns 1 and I. Within subsector reallocation: 4).42 movement of labor from one firm For regressions using labor productivity, to another firm in the same NACE 2 the selection term for surviving firms is digit subsector; positive and significant, implying that surviving is positively correlated with II. Across subsector reallocation: job creation. Notably, if one looks at movement of labor from one firm the first-stage results, the coefficients to another firm in a different NACE on initial productivity in regressions 2 digit subsector but within same predicting whether a firm has been an broad sector (manufacturing, entrant, exiter, or entrant and exiter in services, mining, construction); and the 2005-2010 period are all negative and significant, implying that firms with III. Across sector reallocation: lower productivity are more likely to movement of labor from one firm exit the market and that surviving firms to another firm in a different broad have the highest initial productivity. sector. 42- The analysis was also conducted with quintiles of LP and TFP productivity measures, revealing that with LP , the relationship between initial productivity and job creation holds throughout the productivity distribution (as coefficients on all quintiles are positive and significantly different from the omitted first productivity quintile, as well as rising in magnitude). However, the analysis with TFP as a measure of productivity demonstrates that only firms in the top productivity quintile create more jobs than the rest of firms. 76 GOOD JOBS IN TURKEY Introducing sector fixed effects does and services) are controlled for, the not change the main finding: more relationship between productivity productive firms in Turkey, on average, and job creation rises for both LP and create more jobs (Table 3.5 and TFP , implying a positive contribution Figure 3.25 summarize the results in of across-subsector reallocation Table A3.1.). Even when we control (contributing 46 percent for LP and for detailed subsectors, initial labor 17 percent for TFP). Finally, the productivity is significantly correlated contribution of across-sector with net job creation, implying reallocation (or structural change) can substantial productivity-enhancing, be calculated as the difference between within-subsector reallocation; in the coefficient in regressions with fact, this source of labor reallocation broad sector fixed effects and those accounts for more than half of the without. In the case of TFP , across- total relationship between productivity sector reallocation of labor contributes and job creation (in the case of 13 percent, while it appears that for LP , TFP , the within-subsector reallocation across-sector movement of labor is coefficient is not statistically significant actually slightly productivity-reducing. at the 10 percent level, but it accounts Thus, the results confirm the previous for more than 70 percent of the total finding that within-sector reallocation relationship). When only broad sectors of resources is the most important (i.e., manufacturing, construction, growth-enhancing force in Turkey. TABLE 3.5 The relationship between productivity and labor reallocation in Turkey and its sources, 2005-2010 Coefficients of a cross-sectoral regression Note: ***significant at the 1% level, **significant at the 5% level, *significant at the 10% level. Coefficients can be interpreted as follows: a higher initial labor productivity (LP) results in 0.0807 more labor reallocation when measured within sector(I); 0.065 more when measured across subsector (II) (coefficient on I+II, minus coefficient on I) and -0.003 labor reallocation when measured across sectors (III; i.e. coefficient on I+II+III minus coefficient on I+II). The relative contributions of labor movements in the direction of higher productivity are thus 0.0807/0.143 = 56% for within sector movements, 46% for across subsector, and -2% for across sector movements. Source: Authors’ calculations based on SBS GOOD JOBS IN TURKEY 77 FIGURE 3.25 Sources of productive labor reallocation in Turkey, 2005-2010 Source: Authors’ calculations based on SBS There appears to be labor reallocation productivity with TFP , movement towards more productive activities towards high productivity firms in both within the services sector and manufacturing originates entirely within the manufacturing sector. This from source II: there is movement of is indicated by the positive significant labor from low-productivity to high- coefficients in Table 3.6 (detailed results productivity subsectors, but within are in Table A3.2). However, the source subsectors, there is a slight movement of this positive reallocation differs towards low-productivity firms. The (see Figure 3.26). When measuring result for services is similar across productivity with LP , productive the two measures of productivity: reallocation in manufacturing productive reallocation of labor emanates from both sources I and occurs only within subsectors, with II, with within-subsector reallocation unproductive reallocation across slightly dominating. When measuring subsectors. 78 GOOD JOBS IN TURKEY TABLE 3.6 The relationship between productivity and labor reallocation within sectors and its sources, 2005-2010 Coefficients of a cross-sectoral regression Services Note: ***significant at the 1% level, **significant at the 5% level, *significant at the 10% level. Coefficients can be interpreted as follows: within the manufacturing sectors, a higher initial labor productivity (LP) results in 0.075 more labor reallocation when measured within-sector (I); 0.049 more when measured across subsector (II) (coefficient on I+II, minus coefficient on I). The relative contributions of labor movements in the direction of higher productivity are thus 0.075/0.124 = 60% for within sector movements, and 40% for across subsector. Source: Authors’ calculations based on SBS FIGURE 3.26 Sources of productive labor reallocation in manufacturing and services in Turkey, 2005-2010 Manufacturing: Services: Source: Authors’ calculations based on SBS GOOD JOBS IN TURKEY 79 Panel results the level of previous year’s productivity is controlled, annual growth in the As expected, a productivity increase productivity of a firm is associated with in a firm is associated with a short-run a contraction in the labor resources that downsizing rather than expansion of the firm uses. Although this relationship employment in that firm. As previously between productivity and resources is stated, there is a short-run tradeoff negative for both LP and TFP , it is larger between productivity and job creation: for LP . This result is not surprising since any additional employees hired by LP by definition is more directly tied to the firm necessarily depress the firm’s the number of workers in a firm than productivity in the short run as they is TFP . Notably, the coefficients on the try to catch up to the productivity of previous year’s productivity are all existing employees, so the immediate positive and significant. Thus, although relationship between increases in firms that experience productivity a firm’s productivity and its net job growth shed jobs in the short run, the creation is negative. Panel regressions firms with higher productivity levels confirm that this holds for Turkey as well are still creating more jobs in the next (see Table 3.7 and Table A3.3). Once period than lower-productivity firms. TABLE 3.7 Short-term relationship between productivity growth and job creation Coefficients of a panel regression of change in job creation on change in productivity (percentage points (ppt)) (percentage points (ppt)) Note: ***significant at the 1% level, **significant at the 5% level, *significant at the 10% level. Coefficients can be interpreted as follows: a short term (annual) 1% growth in labor productivity LP results in a 0.055 ppt reduction in the job creation rate. Source: Authors’ calculations based on SBS 80 GOOD JOBS IN TURKEY Decompositions for manufacturing Turkish firm in the short run. The cross sector effect in annual decompositions is a parallel measure to the coefficient Annual productivity decomposition on productivity change in the panel of manufacturing firms supports the regressions. The cross effect takes a previous finding of a negative short- negative sign if expanding firms are term relationship between job creation losing productivity or contracting and productivity growth. The results of firms are experiencing a productivity the FHK decomposition presented in increase. A very high cross effect of Table 3.8 confirm the finding that job -0.81 indicates a strong, negative short- creation and productivity growth move run relationship between job creation in opposite directions for the average and labor productivity. TABLE 3.8 Productivity growth decomposition for manufacturing, 2005-2009, average of annual figures (percentage points (ppt)) Contribution of different kinds of productivity growth to overall productivity growth Note: See Box 3.4 for an explanation of the methodology. Figures can be interpreted as follows: a 1% increase in the productivity of existing firms, weighted by their initial output shares, translates into a 0.77 ppt increase of the overall productivity growth (within effect), a 1% productivity increase achieved through the reallocation of labor from lower to higher productivity firms (between effect) translates into a 0.99 ppt increase in the growth rate of overall productivity. Source: Authors’ calculations based on SBS Recent productivity growth in the Turkey and Latvia. For these two manufacturing sector in Turkey countries, movement of labor originated both from the productivity from low- to high-productivity firms growth experienced by existing was as important (and in the case of firms (“within effect”) and the Turkey, more important) a determinant reallocation of labor from low- to of overall productivity growth as the high-productivity firms (“between within-firm effect. Indeed, Turkey is the effect”). Figure 3.27 compares the country with the highest contribution of results of the FHK decomposition between-firm effect. This confirms our for Turkey’s manufacturing sector in previous findings of substantial creative 2005-2009 period to a few countries destruction happening in Turkey’s from Bartelsman et al. (2004). Within- manufacturing sector. firm productivity growth has been by a large margin the main source Exit of firms has not contributed for overall productivity growth in to productivity growth in Turkey’s all countries with the exception of manufacturing sector. Among the GOOD JOBS IN TURKEY 81 countries in Figure 3.27, Turkey is the made a positive, albeit small (5 percent) only one that experienced a negative contribution to the productivity growth “exit effect” -- a productivity loss due of Turkey’s manufacturing sector, with to the fact that the firms exited the firms that existed in 2009 but not in market over the 2005-2009 period had 2005 being, on average, slightly more been slightly more productive than productive than the average firm in the average firm productivity in 2005. 2005. Relative to comparator countries, Similar to other comparison countries, such “entry effect” is in the middle of Turkey’s “cross effect” is negative the group—above Chile, France, and and of average magnitude, implying the U.S., but below Latvia and Slovenia. that firms experiencing an increase in A positive entry effect might indicate productivity lost employment shares, the existence of high entry barriers in i.e. their productivity growth was Turkey, allowing only firms over a high associated with downsizing rather than productivity threshold to enter the expansion.43 The “entry effect” has market.44 FIGURE 3.27 Productivity growth decomposition in Turkey and selected countries, manufacturing sector (percentage points (ppt)) Note: Data period: Turkey: 2005-2009, Argentina: 1995-2001, Chile: 1985-1999, France: 1990-1995, Latvia: 2001-2002, Portugal: 1991-1994, Slovenia: 1997-2001, U.S.: 1992-1997. Computations make use of firms with a minimum number of 9 employees for Argentina, 20 for Turkey, 10 for Chile, and 1 for all others. Note: See Box 3.4 for an explanation of the methodology. Source: Bartelsman et al. (2004) and authors’ calculations for Turkey 43- Notably, the magnitude of the cross effect decreased from -0.81 in annual calculations to -0.3 in the 5-year differencing, implying that substantial adjustment has occurred to mitigate this negative effect. 44- The fact that Latvia and Slovenia, both transition countries, had the highest entry effect provides support to the argument that the entry effect is correlated with entry barriers. 82 GOOD JOBS IN TURKEY Overall, it appears that there is labor has the lowest labor productivity reallocation in Turkey from less among all sectors, which reduces productive to more productive non- Turkey’s overall labor productivity agricultural firms. The analysis above compared with OECD countries demonstrates that more productive and other emerging markets. Even firms in Turkey appear to be creating though growth-enhancing structural more jobs than less productive firms. change can help raise Turkey’s overall This movement of labor is happening productivity through movement of to a certain extent within both the labor from agriculture to other sectors, manufacturing and services sectors. agriculture will remain an important The movement of workers from sector for years to come. As of 2012, low-productivity activities to high- agriculture contributed more than 9 productivity activities bodes well for percent of GDP and employed close the creation of good jobs and for future to 25 percent of all workers (Figure economic growth in Turkey. 3.28). Thus, in the medium term, it is important to identify the features of 3.4 Are Jobs Being Created in good (i.e., higher-productivity) jobs Agriculture Increasingly More within the agricultural sector, analyze Productive? whether there has been movement to such good jobs within this sector, While agriculture as a whole has and determine strategies to facilitate the lowest labor productivity of such movement in the future. This all economic sectors in Turkey, it section analyzes the recent increase is important to identify good jobs in agricultural employment in Turkey, within this sector given its continued focusing on inter-regional differences importance in terms of output and in job creation and productivity in this employment. Agriculture consistently sector. FIGURE 3.28 Output and employment in agriculture, 1998-2012 Output and employment Change in output and employment Source: Authors’ calculations, based on TUIK. GOOD JOBS IN TURKEY 83 Turkish agricultural employment, increase its share in total employment which had been diminishing since continuously after 2007, and in 2011 the 1980s, reversed course starting it reached 6.1 million (25.5 percent during the economic crisis of 2008-09, of all employed people). While the and continued to increase during the total number of employed people recovery. Between 2004 and 2011, a increased by 26.4 percent (5.19 million U-shaped trend was observed in the people) between 2004 and 2012, the rate of agricultural employment in total contribution of the agriculture sector employment, due to the agricultural to this rise was not noteworthy due to shedding observed till 2007 (Figure the decline between 2004 and 2008. 3.28). In 2004, 29.1 percent of all Agricultural employment increased employed people (5.7 million) were by only 6.7 percent (around 384,000 working in the agricultural sector. people) between 2004 and 2012. Figure While the number of employed people 3.29 shows the annual change in the in Turkey increased between 2004 number of employed people in Turkey and 2012, the number of agricultural as a whole and in the agricultural sector workers decreased from 2004 to 2007, between 2004 and 2012. The decline and the rate of agricultural employment in the share of employment stopped in dipped in 2007 (to 4.8 million people), 2007 and increased until 2011. In 2012, or to 23.5 percent of all employment. the share declined around 1 percent, Agricultural employment started to but is still higher than in 2007. FIGURE 3.29 Annual change in the number of employed people (%) Source: Authors’ calculations, based on TUIK. 84 GOOD JOBS IN TURKEY There is wide inter-regional variation in when agricultural employment in the trend of agricultural employment, Turkey started rising again, it continued and major differences were observed to decline in four NUTS 1 regions. In between regions producing different 2011, agricultural employment fell in crops. Table 3.9 presents the change five NUTS 1 regions, however only one in the agricultural employment in 12 of these regions (West Marmara) was NUTS 1 regions of Turkey (see Figure among the ones that experienced a fall 3.30 for the NUTS 1 map).45 In 2008, in 2008. TABLE 3.9 Annual change of agricultural employment (000’s of people), 2005-2012 Note: Istanbul is excluded due to minimal engagement in agriculture. Source: TUIK. 45- The NUTS (Nomenclature of territorial units for statistics) classification is a hierarchical system for dividing up the economic territory of the EU for the purpose of the collection, development and harmonization of regional statistics, and socio-economic analyses of the regions. Starting from 2002, Turkey has been using NUTS classification. Turkey has 12 NUTS-1 regions, 26 NUTS-2 sub-regions, and 81 NUTS-3 provinces. See annex for the full list of NUTS regions of Turkey. GOOD JOBS IN TURKEY 85 FIGURE 3.30 Map of NUTS 1 regions of Turkey Source: TUIK. The rise in agricultural employment areas and went back to their hometowns is generally explained by reverse - the only place to go. However, analysis migration of urban workers, but it of regional net migration rates (Figure is not supported by inter-regional 3.31) reveals a continuation of migration migration trends. In general, the inflow away from agriculture-intensive regions to agricultural employment during of Turkey from 2007 to 2012, negating the crisis has been explained with the the hypothesis that reverse migration reverse migration of mostly informal has driven the increase in agricultural workers who lost their jobs in urban employment. 86 GOOD JOBS IN TURKEY FIGURE 3.31 Net migration rate and share of agriculture a. in Employment b. in GDP Source: Authors’ calculations based on TUIK. The continued rise in agricultural Focusing on regional variation reveals employment during the economic that the main reason for increased recovery attracted the attention of agricultural employment was the rise many researchers. Using a multi- in agricultural prices in Turkey which sector equilibrium model, Şengül paralleled world prices. Gürsel and and Üngör (2011) associated the İmamoğlu (2013) developed a two- increased employment in agriculture sector small-economy model that with the declining agricultural labor found that the trends in agricultural productivity. Hatunoglu (2011) and employment in 26 NUTS-2 regions the World Bank (2013) argued that the were correlated with the dynamics of dynamics of agricultural employment the prices and production of agricultural increase could be attributed to rising products in the regions. The study agricultural producer prices and salaries also examined the influence of non- of agricultural employees, especially agricultural income, agricultural area between 2007 and 2011. None of use, and agricultural exports on regional these studies addressed the above trends in agricultural employment, mentioned inter-regional variation in but could not establish a significant agricultural employment trends. relationship. GOOD JOBS IN TURKEY 87 Between 2007 and 2009, Turkey therefore, agricultural productivity, significantly changed the support unfortunately so far, data on regional scheme for its agricultural sector, producer support estimates (including moving to a product-specific and the market price support) and regional production-dependent support subsidies in NUTS-2 regions have not scheme, and part of the change in been made available, thus preventing production could be related with any further analysis of the relationship that change. Starting in 2001, the between agricultural support schemes, Turkish agricultural support system productivity, and job creation. was replaced with a direct income support program, with the transfers The rise in agricultural employment to producers decoupled from actual between 2006 and 2012 seems to be production and paid in line with the concentrated in the regions with area of cultivated agricultural land. higher labor productivity. There is a The aim of that reform was to minimize positive correlation between labor production distortions caused by the productivity in agriculture –calculated previous subsidy scheme. However, as agricultural GDP per worker— the reform was never fully implemented in 2006 and the rise in the share of and state intervention continued agriculture in total employment through direct purchases of crops, between 2006 and 2012. A similar premium payments for the products correlation exists between regional facing shortages, and high taxation migration rates and agricultural labor of certain products at the borders. In productivity. Hence, there seems to 2005, the support system was revisited be a reallocation of labor in agriculture and the amounts of production-related towards more productive regions payments were increased gradually together with a rise in agriculture until 2009, when the transformation intensity in these regions, which might was fully complete. Although the be considered a sign of increasing extent of agricultural support affects employment in more productive jobs in production and employment, and agriculture. 88 GOOD JOBS IN TURKEY FIGURE 3.32 Regional productivity, employment and migration in agriculture, 2006-2012 a. Change in agricultural employment share, 2006-2012 b. Migration (percent) (percent) Source: Authors’ calculations based on TUIK. 3.5 Conclusion and Policy enhancing, but can still be accelerated. Outlook Regression analysis demonstrates that there is, in fact, labor reallocation Overall, labor reallocation in Turkey in Turkey from lower- to higher- has been growth-enhancing. The productivity non-agricultural firms. strongest support to overall growth This movement of labor is happening has come from the movement from to a certain extent within both the agriculture to non-agriculture. Indeed, manufacturing and the services sectors the employment share of agriculture as well as between the two sectors fell from 41.5 percent in 1998 to 24.5 and subsectors within them. However, percent in 2012. As agriculture had the there is also limited evidence of labor lowest labor productivity of all sectors in movements that diminish growth. 2011, movements away from this sector are reflected as improving productivity. Labor reallocation in agriculture appears to be growth-enhancing, even Labor movements within the non- though the sector as a whole has low agricultural sectors are overall growth- labor productivity. Descriptive evidence GOOD JOBS IN TURKEY 89 supports that labor movement has been to have borne fruit already, as towards more productive activities. The suggested by the significant entry main finding is that regions with higher of 20-49 employee firms in the 2010 agricultural productivity (such as the SBS firm census. Mediterranean and South East Anatolia) appear to have increased their shares • Initiatives expanding the scope of agricultural employment, while those for flexible contracting under that have features correlated with lower consideration in the National productivity (such as East Anatolia and Employment Strategy can facilitate Black Sea) have experienced declines labor mobility and reallocation in the agricultural share of employment. without jeopardizing workers’ While this suggests some reallocation security. of labor within agriculture towards more productive activities, more work needs • The reform of severance pay, to be done on this subject, incorporating which is very high by international the influence of agricultural support standards and might be reducing schemes and identifying best practices productive labor reallocation, is one that can be applied to stimulate the of the policies envisioned in the 10th creation of good jobs within agriculture Development Plan. in Turkey. • In agriculture, the availability of a The Government of Turkey has already public support scheme appears to adopted or is considering measures that have had a supportive effect on can accelerate the movement of labor formalization and reallocation of towards more productive activities: labor to more productive regions. • Initiatives encouraging employment of women and youth through Productivity and formality of a job go reductions in the employer share of hand-in-hand. The government can social security taxes can potentially possibly lower the cost of formality facilitate the integration of existing by ensuring or strengthening the local rural-to-urban migrants into presence of services involved in tax and productive activities and accelerate social security benefit administration. such mobility, and with it agricultural shedding, in the future. The quality of a job, including formality, is important for livelihoods. This topic • Implementation of the action plan takes us to the next chapter, which on combating informality appears explores jobs and living standards. 90 GOOD JOBS IN TURKEY Chapter 3 References Aldan, M. Can, and Erol Cakmak, 2011, Hallward-Driemeier, Mary and Reyes “Agricultural Employment in Turkey Aterido. 2013. “Who Creates Jobs? and Regional Differences.” Turkonfed Differences in Firm Dynamics Across Turkpres 2011 Symposium Book, pp. Developing Countries.” World Bank 77-100. mimeo Ayyagari, Meghana, Asli Demirguc-Kunt, Hatunoğlu, E. E. 2011. “Developments and Vojislav Maksimovic. 2011. “Small in Agricultural Sector Employment.” vs. Young Firms across the World: Background paper for World Bank and Contribution to Employment, Job Ministry of Development 2013. “Turkey: Creation, and Growth.” World Bank Managing Labor Markets through Policy Research Working Paper 5631. the Economic Cycle.” World Bank, World Bank, Washington, DC. Washington, DC. Bartelsman, Eric, John Haltiwanger, McMillan, Margaret, and Dani Rodrik. 2011. and Stefano Scarpetta. 2004. “Globalization, Structural Change and “Microeconomic Evidence of Creative Productivity Growth.” NBER Working Destruction in Industrial and Developing Paper 17143. Countries.” IZA Discussion Paper No. Rodrik, Dani, 2010. “Structural 1374. Transformation and Economic Comola, M., and L. de Mello. 2009. “The Development.” Draft paper delivered as Determinants of Employment and the Merih Celasun Memorial Lecture at Earnings in Indonesia: A Multinomial TEPAV, Ankara, on December 22, 2010. Selection Approach.” OECD Economics Şengül, Gonul, and Murat Üngör. 2011. Department Working Papers, No. 690, “Increasing Share of Agriculture in OECD Publishing. Employment in the Time of Crises: Davis, Steven, and John Haltiwanger. 1999. Puzzle or not?” Central Bank of Turkey “Gross Job Flows.” In Handbook of Working Paper No:11/05. Labor Economics, ed. O. Ashenfelter World Bank. 2012. World Development and D. Card, Edition 1, Volume 3, Report: Jobs (overview). Washington, Chapter 41, pp. 2711-2805. DC: The World Bank. Gürsel, Seyfettin, and Zümrüt Imamoglu. World Bank and Ministry of Development 2013. “Why is Agricultural Employment 2013. “Turkey: Managing Labor Markets Increasing in Turkey?” BETAM Working through the Economic Cycle.” World Paper Series #004. Bank, Washington, DC. GOOD JOBS IN TURKEY 91 Annex 3.1: Additional Figures and Tables FIGURE A3.1 Job creation rate distributions, 2010 By sector By firm size Source: Author’s calculations based on SBS 2010 92 GOOD JOBS IN TURKEY FIGURE A3.2 Productivity distributions, 2010 Labor Productivity TFP a) By Sector b) By Firm Type c) By Firm Size d) By Region Note: Computation based on 20+ firms only. Source: Author’s calculations based on SBS GOOD JOBS IN TURKEY 93 TABLE A3.1 Productivity and Net Job Creation in Turkey, 2005-2010 Note: Multinomial logit regression, with second-stage dependent variable of net job creation rate during the period of 2005 and 2010 for surviving firms; exporter status (=1 if exporter, 0 otherwise) included in the second-stage regression. First-stage predicts the probabilities of being an entrant, exiter, and entrant &exiter firm using initial productivity (either LP or TFP), exporter status and NUTS2 region dummies (as the excluded variables from the second stage). *** p<0.01, ** p<0.05, * p<0.1. Source: Author’s calculations based on SBS 94 GOOD JOBS IN TURKEY TABLE A3.2 Productivity and Net Job Creation in Manufacturing and Services in Turkey, 2005-2010 Notes: Multinomial logit regression, with second-stage dependent variable of net job creation rate during the period of 2005 and 2010 for surviving firms; exporter status (=1 if exporter, 0 otherwise) included in the second-stage regression. First-stage predicts the probabilities of being an entrant, exiter, and entrant &exiter firm using initial productivity (either LP or TFP), exporter status and NUTS2 region dummies (as the excluded variables from the second stage). *** p<0.01, ** p<0.05, * p<0.1. GOOD JOBS IN TURKEY 95 TABLE A3.3 Productivity, Productivity Growth, and Net Job Creation in Turkey, 2005-2010 *** p<0.01, ** p<0.05, * p<0.1. 96 GOOD JOBS IN TURKEY GOOD JOBS IN TURKEY 97 4. Jobs and Living Standards in Turkey Abstract: Turkey’s recent growth was the employment opportunities largely inclusive, but little attention has of underprivileged groups (e.g., been paid to the types of jobs generated youth, women, and the long-term and their contribution to higher welfare. unemployed) should be enhanced to To better understand the role of jobs promote living standards among low- for raising welfare levels, trends in income households. living standards among in-work and out-of-work households are analyzed. 4.1 Introduction Two conceptually different indicators of living standards are considered: As noted in the introduction to this the low-income rate and material report, Turkey’s recent growth has been deprivation. Living standards among largely inclusive, as the welfare of the in-work households remained largely poorest 40 percent of the population unchanged; at the same time, the grew faster than the average. Better absolute number of in-work households labor market performance certainly increased significantly, suggesting that contributed to the higher welfare level the strong economic growth enabled of the poorest, and women in particular more people to successfully participate benefitted from these positive trends. in the labor market and enjoy higher Since 2005, female labor force welfare levels. Living standards were participation and employment rates positively associated with different have steadily increased (recall Chapter types of employment, but mainly for 2). Yet, despite these improvements, non-agricultural jobs. Furthermore, less than half of the working-age labor income was the biggest population (WAP) is employed in 2012 contributor to total household income and 39 percent of workers are still and growth in labor income itself had informal (World Bank 2013). a positive impact on living standards among low-income households. To In the ample literature on employment promote higher living standards, in Turkey, little attention has been paid public policy should improve access to so far to the types of jobs generated and quality of education and training and their contribution to higher welfare programs as there is a strong link levels. The employment surge was between the educational level and mainly in the service sector and in earnings potential of workers. Similarly, formal employment, and primarily 98 GOOD JOBS IN TURKEY benefited college graduates, although on Turkey’s Survey on Income and informal agricultural employment also Living Conditions (SILC) for survey increased among women (Chapter 2). years 2006-2010. Most employment- The type of job (for example, salaried related data, including income, refer to formal work, casual or seasonal work, the respective previous calendar year or self-employment) can be linked with (the reference year). As the focus was the income and welfare of the job holder, on the contribution of jobs to raising as well as those of his household. living standards, the sample included all households with at least one member Chapter 4 focuses on the role of jobs of working age (15-64 years old).46 If in raising living standards. To fully at least one adult member was working appreciate the dynamics among in- full-time47 during the entire reference work and out-of-work households, the year, the household was categorized as analysis first looks at overall trends “in-work.”48 Otherwise, the household among households with working-age was considered “out-of-work.” members. The chapter then documents levels of and trends in living standards To provide a comprehensive analysis in Turkey for the period 2006-2010. of the levels of and trends in living Analyzing the determinants of welfare standards in Turkey, two conceptually sheds more light onto the role that jobs different indicators were applied: have played in raising living standards. (i) the rate of low income;and (ii) an In a final step, changes in living index of material deprivation. The low- standards are decomposed to identify income threshold is defined as the the individual contributions of labor income level of the bottom 10 percent income, non-labor income, and altered of the population in 2006, and is held household composition. constant over time. All incomes were deflated using the national consumer 4.1.1 Methodology price index (CPI) and expressed in 2006 prices. For most parts of the analysis, The analysis in this chapter was based total disposable household income was 46- A significant number of older workers are still “in-work,” especially in rural areas. In 2006, around 8 percent of people aged 65 and above in rural areas had been in full-time employment for at least one month during the reference year. 47- To group together households with comparable employment states, time spent in part-time jobs was not considered. The incidence of part-time employment varied between 4 percent in urban areas and 13 percent in rural areas in 2006. 48- See Annex 4.1 Figure A4.1 for more details. To test the sensitivity of the results, working full-time for at least six months during the reference year was used as an alternative cut-off point. Results are largely the same and available upon request. GOOD JOBS IN TURKEY 99 adjusted to account for differences in cells. NUTS1 regions were aggregated household size and composition using into Western and Eastern clusters.49 the OECD equivalence scale. Within each cluster, rural and urban areas were identified using the SILC Material deprivation takes a more data. This approach allowed for a better direct approach to measuring living illustration of the different dynamics standards. For the analysis that follows, between the Eastern and Western parts households that could not satisfy three of Turkey, as well as between rural basic needs were considered materially and urban areas. Second, results are deprived: notably, to provide food, typically presented at the individual clothing, and heating. In particular, level. To allow an easier interpretation, the household head was asked about: the discussion sometimes refers to the (i) the capacity to afford meals with total number of households (instead of meat, chicken, or fish (or vegetarian total population). equivalent) every second day; (ii) the capacity to replace worn out clothes by Before presenting the results, the next new ones; and (iii) the ability to keep section briefly discusses recent trends the home adequately warm. This way, regarding the number of households the welfare level is measured directly with working-age members and the through the consumption of certain labor market attachment of these goods – instead of indirectly through households, important background disposable income. Since monetary information for the analysis that and non-monetary indicators of well- follows. being are often only weakly correlated, outcome-based measures, such as 4.1.2 Dynamics at the household the material deprivation index used level here, add important information to the discussion of raising living standards. The total number of households with at least one member of working Results are presented in several ways. age rose by 10.2 percent from 16.1 First, to keep the discussion clear while million in 2006 to 17.8 million in 2010. acknowledging the heterogeneity The annual increase in the number of of the Turkish economy, regions are households with working-age members grouped into four distinct geographic was particularly high in 2009, at 5.3 49- The Western cluster includes Istanbul, West and East Marmara, Aegean, West and Central Anatolia, Mediterranean, and West Black Sea. The Eastern cluster contains East Black Sea, Northeast Anatolia, Middle East Anatolia, and Southeast Anatolia. 100 GOOD JOBS IN TURKEY percent. Regional growth rates also The increases were biggest in the peaked in 2009, except in West-Rural, Eastern provinces, where traditionally where growth was highest in 2008 (13.6 fewer households are categorized percent). Over the entire period, the as “in-work.” Growth in the Western increases were biggest in rural areas provinces was more moderate, resulting (16.5 percent). in smaller regional gaps (Figure 4.1). In 2010, the share of in-work households The share of in-work households among was highest in East-Rural (70.3 percent), all households 50 increased slightly from followed by West-Rural (69.4 percent), 66.3 percent to 68.4 percent, but there West-Urban (68.1 percent), and East- were large differences across regions. Urban (66.5 percent). FIGURE 4.1 Share of in-work households, nationally and by region, 2006-2010 Note: Unit of observation is the household. Source: SILC. 50- “All households” refers to the total number of households with at least one member of working age (15-64 years old). Please note that the information to discriminate between in-work households and out-of-work households referred to the previous calendar year (reference year). GOOD JOBS IN TURKEY 101 The share of in-work households the economic situation of households peaked in 2009, mainly the result of two with at least one member of working factors. In survey year 2009: (i) the total age (15-64 years). To shed more light number of households with working- onto the role of jobs in raising welfare age members increased significantly; levels, trends in living standards among and (ii) the share of people reporting in-work and out-of-work households that they had worked full-time for 12 are reported. In a first step, the share months during the previous calendar of the population living in low-income year increased considerably (Annex households is analyzed; secondly, the 4.1 Figure A4.1). As a result of these rate of material deprivation among two factors, the increase in in-work the Turkish population is examined. households was above average in Material deprivation measures the 2009. ability of households to satisfy three basis needs related to food, clothing, The share of households with no regular and heating.51 labor market attachment increased sharply during the crisis. Unemployment 4.2.1 The incidence of low income rose significantly from 10.9 percent in 2008 to 14.0 percent in 2009, especially Despite strong economic growth prior in urban areas (IMF 2013; World Bank to the crisis, the share of the population 2011). The share of people working living in low-income households full-time for 12 months also declined increased slightly. The share of the during 2009 (Annex 4.1 Figure A4.1). population living in households with Both factors left more households with low incomes rose slightly from 10.2 no regular labor market attachment and percent in 2007 to 11.7 percent in 2010. the share of out-of-work households However, national averages mask large increased considerably. East-Urban differences across regions and by a reas were particularly affected; household type (Figure 4.2). The share the share of out-of-work households was lower for in-work households, rose from 16.9 percent to 33.5 percent especially in urban areas and Western (East-Rural areas: 16.3 percent to 29.7 provinces. Between 2007 and 2010, percent). living standards remained relatively constant; the share of the population 4.2 Levels of and Trends in below the income threshold increased Living Standards by around 1 percentage point. In-work households in East-Rural experienced The following section briefly describes more difficulty, and the share of people 51- Trends in material deprivation were analyzed over the period 2006-2010. As there were concerns regarding the income distribution in 2005 (reported in survey year 2006), trends in the low-income rate were analyzed for the years 2006-2009 only (using survey years 2007-2010). See Annex Figure A4.2 for details. 102 GOOD JOBS IN TURKEY below the threshold increased from percent.52 23.6 percent in 2007 to 29.3 percent in 2010. Among households without Turkey was hit hard by the crisis; its regular labor market attachment, impact on living standards differed regional differences were even larger across regions and household types. and increased over time, mainly driven The main transmission mechanism of by opposite trends between urban and the crisis was the labor market. Higher rural areas. In 2010, the share of people unemployment rates in 2009, especially in low-income households ranged in urban areas, contributed to the larger from 9.5 percent in West-Urban to 62.8 number of households with no regular percent in East-Rural. labor market attachment. Also, out- of-work households’ income declined The gap in living standards between (Annex 4.1 Table A4.1). To support household types became smaller, people out of work, unemployment mainly driven by an increasing share of benefits have increased in 2009 (as in-work households with low income. were minimum pensions) which Between 2007 and 2010, the share of certainly helped vulnerable households out-of-work households below the better cope with the crisis. Yet for these income threshold decreased somewhat, households the share living below the from 23.1 percent to 22.1 percent. income threshold rose significantly, As the share of in-work, low-income but stayed slightly below the pre-crisis households increased from 5.6 percent level. in 2007 to 7.6 percent in 2010, the gap decreased from 17.4 percent to 14.5 To combat the crisis, a number of percent in 2010. The gap was smallest policy measures were implemented in 2009, related to the disproportionate to keep people in the labor market. decrease in out-of-work households. For example, legislation on short work The compositional shift also affected was defined better and made more the distribution of income; out-of-work generous and accessible53; and social households reported higher incomes security contributions were lowered in 2009, and the share of people in both across the board and for youth low-income households declined and women in particular. In-work considerably. The opposite trend households, especially at the lower occurred among in-work households, end of the income distribution, also and the proportion of people below benefited from annual increases of the the income threshold increased to 11.3 real minimum wages. 52- For more details regarding income levels for both types of households, see Annex 4.1 Table A4.1. 53- For example, a statutory change authorized by law 5383, and subsequent publications in Official Gazette No. 2711, including an extension of the maximum time of short time work subsidy payment. GOOD JOBS IN TURKEY 103 FIGURE 4.2 Population share living in low-income households, 2007-2010 Note: Unit of observation is the individual. Source: SILC. 4.2.2 Material deprivation be deprived of certain basic goods than in-work households, but the gap has Material deprivation is still widespread become much smaller. Among in-work among Turkish households, but households, the material deprivation significant improvements have been rate remained relatively constant until achieved since 2006. At the national 2009; in 2010, it declined considerably, level, the share of people without reaching 17.6 percent. In contrast, for access to basic goods declined from 29 out-of-work households, the situation percent in 2006 to 21 percent in 2010. has steadily improved over time; Improvements were uniform across progress has been particularly strong the population; material deprivation was lower in 2010 than in 2006 for both among rural households in Eastern household types and in all regions. provinces, where the share of the materially deprived population declined Households without regular labor from 69 percent in 2006 to 45 percent in market attachment are still more likely to 2010 (Figure 4.3). 104 GOOD JOBS IN TURKEY FIGURE 4.3 Population share affected by material deprivation, 2006-2010 Note: Unit of observation is the individual. Source: SILC. The global financial crisis had and stayed below pre-crisis levels in all only limited impacts on trends in regions. material deprivation. Among in- work households, the share of people 4.2.3 Discussion with unmet basic needs increased somewhat in 2009, whereas for out- Despite differences in levels and of-work households it continued to trends over time, a number of findings decline during the crisis. Again, the hold for both measures of well-being. fact that the number of out-of-work First, in-work households enjoyed households declined by nearly 1.5 higher living standards than out- million in 2009 (but went up in the of-work households, indicating that following year) may have been the jobs contributed to higher levels of main driver of this result. In 2010, welfare. Second, the gap in living material deprivation rates declined standards between in-work and out- significantly for in-work households of-work households decreased, but GOOD JOBS IN TURKEY 105 the underlying reasons differed across standard often also showed high rates the welfare measures. In the case of of material deprivation. This apparent material deprivation, the declining correlation at the regional level trend among out-of-work households was not reflected at the household helped narrow the gap, suggesting level, however. In fact, the overlap that recent trends in non-labor income between the two indicators was rather played an important role in raising small. In survey year 2007, only 6 living standards. Third, living standards percent of all households were both among in-work households remained materially deprived and in the bottom largely unchanged. At the same income decile.54 Materially-deprived time, the absolute number of in-work households can be found in any income households increased significantly, decile (Figure 4.4).55 This suggests that suggesting that the strong economic the underlying reasons for households growth allowed more people to to be below the income threshold or successfully participate in the labor to be materially deprived differ; this, market. in turn, has important implications for policy makers. The following section In Turkey, regions with higher shares sheds more light on the drivers of living of people below the minimum living standards in Turkey. 54- Note that the income data refer to the previous calendar year, and material deprivation is determined using information from the actual survey year. This discrepancy may contribute to a lower correlation between the two indicators. Results are very similar when the analysis is repeated using panel data that allow for the same reference period for both indicators, but do not provide regional information or population weights. The correlation is somewhat stronger for rural households and households with no regular labor market attachment. 55- Similar results were found for many countries. In the EU member states, for example, the population above the poverty line was consistently less affected by material deprivation than the poor, but the respective rates of material deprivation were positive in all countries and varied considerably across countries (Eurostat 2010). 106 GOOD JOBS IN TURKEY FIGURE 4.4 Material deprivation by income decile, 2007 Note: Unit of observation is the household. Source: SILC. 4.3 Determinants of Living below the income threshold; and (ii) Standards suffered from material deprivation. Complete results are reported in Annex To take a more detailed look at the 4.1 Table A4.2. contribution of jobs to higher living standards, we performed a regression The drivers of low income and material analysis on the sample of in-work deprivation in Turkey were similar: age, households. In particular, the personal education level, and years of experience and employment-related information of the breadwinner mattered. A higher of the household’s breadwinner were education level for the breadwinner was used to determine the probability that associated with higher living standards. the breadwinner’s household: (i) lived For example, if the main household GOOD JOBS IN TURKEY 107 earner had graduated from vocational standard. Informal jobs contributed to or technical high school, the chance higher living standards, but only in the that the breadwinner’s household was non-agricultural sectors. An informal not materially deprived increased by 17 agricultural job significantly increased percent.56 In case of completed tertiary the probability of living below the education, it increased by 23 percent. income threshold. Also, self-employed Regarding low-income households, agricultural workers faced lower levels the link between education and living of well-being. This result is consistent standards was weaker – but still highly with the findings of Chapter 3, which significant. If the breadwinner had documented that a move of jobs away completed high school, the chance of from agriculture was associated with living above the minimum welfare level higher productivity in the new job. increased by 1.4 percent; in case of tertiary education, it increased by 1.8 The probability of attaining higher percent. Similarly, the older the head living standards varied considerably of household and the more time s/ by geographical location. Compared he spent in paid work, the higher the to households located in Anatolia, chances that the entire household was households in all other regions found better off. it easier to move up in the income distribution and were less affected by The type of employment of the material deprivation. Also, whether breadwinner mattered for the a household was located in an urban household’s living standard. Higher or rural area played an important living standards were associated with role for living standards. In line with a number of job types, but mainly expectations, the urban population those in the non-agricultural sectors. was less likely to live below the income For example, formal non-agricultural threshold. More and better jobs as well wage employment increased the as higher pay typically lead to smaller household’s chances of moving out of numbers of low-income households in the bottom decile by around 6 percent; urban areas. For Turkey, the effect was the effect was of similar magnitude for small but significant. Regarding material material deprivation. Similarly, when deprivation, the opposite seemed to be the breadwinner was an employer true: urban households faced a higher outside agriculture, his household was risk of material deprivation than rural more likely to enjoy a higher living households.57 Differences in prices for 56- The reference category is no schooling. See also Annex 4.1 Table A4.2. 57- This result is robust across years and various specifications of the regression model. 108 GOOD JOBS IN TURKEY consumer goods, which make it harder particular, variables that better capture for urban households to satisfy basic the specific needs of households (e.g., needs, may partly explain this result. material obligations resulting from Rural households produce most of the household size and structure, marital food they consume, increasing their status, and key life events, as well as capacity to afford regular protein-rich health status of household members meals. Also, rural households may and housing conditions) and a measure benefit more from housing allowances, of the permanent income have been which in turn may help them keep the shown to be strongly linked with material house adequately warm. deprivation (Berthoud and Bryan 2010; Whelan, Layte, and Maître 2004). Households with higher shares of working adults or pensioners 4.4 A More Detailed Look enjoyed higher living standards. at the Trends in Household Adding information regarding the Income demographic composition of the household increased the explanatory Despite strong economic growth, power of the models. A higher share of living standards among low-income working adults significantly increased households have not improved. In the chance of living above the income part, compositional factors may be threshold; it also considerably lowered at play. The number of households the probability of being materially with working-age members increased deprived. In addition, benefits from considerably; many of these “new” old-age pensions seem to matter for households were absorbed into local welfare, as a higher share of pensioners labor markets as the number of in- in the household contributed to higher work households increased as well. living standards. New labor market entrants may have accepted entry-level jobs with relatively To conclude, whether or not a low pay; as the propensity of low household is low-income or materially- income decreases with more labor deprived depends on similar factors. market experience of the breadwinner, Holding a formal job is a crucial these households have a good chance determinant for raising living standards. of moving out of the bottom decile in To better understand what makes a the future. According to the findings household materially deprived despite of Chapter 2, youth have overall less sufficient income, further and more access to formal jobs than older age detailed analyses are needed. In cohorts. Another reason of stagnant GOOD JOBS IN TURKEY 109 living standards may be that most important over time. Their share in total of the years reflected in the analysis household income increased from 15.9 witnessed a move of some employment percent to 17.7 percent (Figure 4.5). back to agriculture. On the other hand, stagnant incomes at the aggregate In rural areas, income from self- level may mask the more dynamic but employment remained the biggest neutralizing trends of the individual contributor to total household income, components of household income. In but its share declined between 2007 and the following section, the main sources 2010. Income from self-employment of total household income are analyzed, was the main income source for with a focus on trends in labor and rural households, but other sources, non-labor income. In a second step, including wage employment and social changes in real per capita household transfers, became more important income are decomposed to highlight over time. For example, among rural the contribution of several factors, households in Eastern provinces, the including labor income, to the observed share of labor income increased from changes in living standards.58 24.8 percent in 2007 to 31.3 percent in 2009 (Annex 4.1 Figure A4.3). As the 4.4.1 Sources of household income decline in self-employment was more pronounced, total labor income became At the national level, the composition of a less important source of household household income changed somewhat over time; labor income and social income; the share declined from 65.9 transfers became more important. Labor percent in 2007 to 63.2 percent in 2010. income from wage employment was Rising unemployment and reduced the main contributor to total household real hourly wages in 2009 contributed income; together with the income from to a declining share of wage income; self-employment, the share increased rural households reverted to self- from 58.6 percent in 2007 to 61.5 employment and informal employment percent in 2010. Social transfers, which as a coping mechanism during the crisis include pensions and unemployment (World Bank 2011 and Chapter 2 of this benefits, were the largest component report). Similar trends were observed for of non-labor income and became more rural households in Western provinces. 58- For this part of the analysis, all households are considered. For more details regarding the decomposition method, please see Azevedo, Sanfelice and Nguyen (2012). 110 GOOD JOBS IN TURKEY FIGURE 4.5 Sources of household income, 2007-2010 Source: SILC. The contribution of labor income to for the first four quintiles, and jumped to total household income was higher almost 23 percent for the top quintile. for richer households. In 2010, for As a result, more than two-thirds of households in the bottom quintile of total household income of households the income distribution, the share of in the top quintile originated from total wage employment was 33 percent; it labor income (wage employment plus increased across quintiles, reaching 46 percent in quintile 4 (Figure 4.6). The self-employment) compared with only contribution of income obtained from 48 percent for households in the bottom self-employment was around 15 percent quintile.59 59- Similar trends were observed in earlier years. In Figure 4.6, pensions have been separated from other social benefits; the share of social transfers declined with household income and contributed 1.7 percent to total household income among households in the top quintile. GOOD JOBS IN TURKEY 111 FIGURE 4.6 Sources of household income by quintile, 2010 Source: SILC. 4.4.2 Decomposing the changes in in Figure 4.7 (right axis), income per living standards capita declined between 2 percent and 3 percent for most deciles. The observed changes in real per capita Middle-income households were household income were decomposed somewhat less affected by the decline. into changes in different types of Households in the top decile of the income and changes in household income distribution realized the largest composition. This approach helps to reduction; real per capita income was highlight the role of employment in around 10 percent lower in 2010 than raising living standards. Given the data in 2007. at hand, the decomposition used cross- sectional data (rather than panel data) Changes in the household structure and analyzed the changes between (share of working-age adults and share 2007 and 2010. Results are presented of workers) contributed to higher living in Figure 4.7. standards. Changes in the composition of households had a positive impact Across all income deciles, real per on living standards (Figure 4.7 left capita household income declined axis). In particular, the share of workers between 2007 and 2010. As illustrated contributed consistently to higher 112 GOOD JOBS IN TURKEY incomes; the effect was of similar size Changes in non-labor income had for most deciles and is consistent with a large negative impact on living the evidence obtained from the analysis standards, offsetting the income- regarding the determinants of living increasing trends of other factors. As standards. shown in the previous section, pensions and other social transfers (including Labor income per working adult had a benefits for unemployment, disability, positive impact on welfare only for the and sickness as well as housing and bottom 10 percent of all households. family allowances) are the main sources Between 2007 and 2010, labor income of non-labor income. Both types of per worker increased by 6.3 percent transfers had small positive effects among low-income households. For all other income deciles, the contribution on the change in per capita income, of labor income was negative. Higher especially for low-income households. minimum wages and crisis-related The large negative impact can be policy interventions (for example, attributed to changes in the unspecified reduced hourly wages) are likely to part of non-labor income (labeled have contributed to these trends. “other” in Figure 4.6). FIGURE 4.7 The contribution of labor income, non-labor income, and household structure to changes in living standards between 2007 and 2010 Note: The columns show the contribution of each component (in percent) to the change in per capita income in each decile. Results are obtained from cross-sectional data. Source: SILC. GOOD JOBS IN TURKEY 113 To conclude, growth in labor income Despite these overall positive trends, has contributed to higher welfare levels living standards have not improved by of low-income households in Turkey, much. A number of reasons are likely to highlighting the important role that have contributed to this result. First, as jobs play in raising living standards. the number of in-work households rose Crisis-related distortions in the labor from 11.7 million in 2007 to 12.2 million market probably added to the difficulty in 2010, the composition within this in identifying longer-term trends on group of households changed. Some changes in living standards. To gain low-income households left the bottom a better understanding regarding decile for more productive and often the contribution of specific income formal jobs, but were “replaced” by new households entering the labor market at sources, longer time series and the lower end of the income distribution. longitudinal data are required. At the aggregate level, the share of low-income households remained 4.5 Conclusion and Policy relatively stagnant. At the individual Outlook level, however, mobility is likely to be higher as: (i) in-work households with For a large majority of people around older, more experienced heads were the world, work is their main source of more likely to move out of the bottom income. Growth in labor income has decile; and (ii) the welfare level of the been the most important contributor “new” in-work households was on to poverty reduction and higher average higher than among out-of-work living standards in many countries households, especially as the overall (World Bank 2012). In Turkey, strong quality of jobs created after the crisis economic growth allowed more was unusually high (see Chapter 2). people to successfully participate in the labor market; the share of in-work Second, the economic sector and households increased significantly and type of employment are critical for in 2010, more than 70 percent of the raising living standards. Between 2000 population lived in households that and 2008, the employment share in were well integrated in the labor market. agriculture declined by 11 percentage Labor income (wage employment plus points; employment shares in the earnings from self-employment) was the trade, services, and manufacturing biggest contributor to total household sectors increased considerably, but income; for urban households, its share could not replace all the jobs that has increased in recent years. Among were lost in agriculture (ILO 2011). The low-income households, growth in structural labor reallocation process labor income per worker was positive from agriculture to sectors with higher between 2007 and 2010 and contributed labor productivity is critical for raising to higher living standards. living standards. However, it also 114 GOOD JOBS IN TURKEY matters what type of jobs are created in of the most pressing issues of the the non-agricultural sectors. The share Turkish economy (Özkan 2012). One of informal jobs declined somewhat central pillar of the NES is increasing during the period of economic growth, educational outcomes and improving but casual wage employment increased training opportunities. As shown in this significantly. The share of low-income report and in the related literature, there households where the breadwinner is a strong link between the educational was casually employed was around 15 level and earnings potential of workers. percent in 2007 and 2008, and jumped Public policy can remove education- to 25 percent in 2009.60 Many of these related barriers to income mobility by: casual workers found employment (i) increasing access to and quality of in the non-agricultural sector, but the education for adults and children; (ii) low wages associated with this type implementing redistributive policies; of employment have not led to higher and (iii) developing social transfer living standards. programs (OECD 2010). In particular, adult training programs allow current Third, despite its limited impact on the workers to improve their set of Turkish economy, the global financial skills, giving them access to higher crisis may have adversely affected the productivity jobs with better pay. mobility of low-income households. Unemployment rose sharply, especially As a second pillar of the NES, in urban areas; informal workers were employment opportunities for disproportionally affected both by underprivileged groups such as job losses and wage reductions (ILO youth, women, and the long-term 2011). Formal wage earners were more unemployed should be promoted; this likely to keep their jobs, but as part in turn is likely to reduce the share of of the policy interventions during the low-income households. Given the low crisis, real hourly wages were reduced earnings potential for some of these significantly (World Bank 2011) before groups, a comprehensive activation the minimum wage was increased. of cash transfers could help raise the human capital and living standards of Besides developing an environment the entire household. For households that allows firms to grow and with no regular labor market attachment, create more and better jobs, policy social transfers have been shown to makers could build on the post-crisis play an important role in raising living achievements and help reduce the standards. Improved targeting and share of low-income households. social transfers that also take into In 2012, the Government of Turkey consideration regional differences may adopted a National Employment help to further increase the well-being Strategy (NES) that addresses some of out-of-work households. 60- It dropped to 14 percent in 2010. GOOD JOBS IN TURKEY 115 Chapter 4 References Azevedo, J. P., V. Sanfelice, and M. Nguyen. Özkan, M. A. 2012. “The Turkish Labor 2012. “Shapley decomposition by Market.” Speech delivered at the components of a welfare aggregate.” ETF Regional Conference “Social World Bank mimeo. World Bank, Partnership in VET in the Southern and Washington, DC. Eastern Mediterranean Region, Istanbul, Oktober 2012. Berthoud, R., and M. Bryan. 2010. “Income, deprivation and poverty: a longitudinal Whelan, C. T., R. Layte, and B. Maître. 2004. analysis.” Journal of Social Policy Vol. “Understanding the mismatch between 40, No. 1, pp. 135-156. income poverty and deprivation: a dynamic comparative analysis.” Eurostat. 2010. “Income poverty and European Sociological Review Vol. 20, material deprivation in European No. 4, pp. 287-302. countries.” Eurostat – Methodologies World Bank. 2011. “The jobs crisis: and Working Papers. Household and government responses ILO. 2011. “Occupational Outlook in Turkey.” to the great Recession in Eastern International Labour Organization: Europe and Central Asia.” World Bank, Ankara. Washington, DC. IMF. 2013. “World Economic Outlook.” World Bank. 2012. World Development April. International Monetary Fund, Report 2013: Jobs. Washington, DC: Washington, DC. World Bank. OECD. 2010. “Economic Policy Reform: World Bank. 2013. “Impact study - Turkey: Going for growth.” Organisation Evaluating the Impact of İŞKUR’s for Economic Co-operation and Vocational Training Programs.” World Development, Paris. Bank, Washington, DC. 116 GOOD JOBS IN TURKEY Annex 4.1: Background Figures and Tables FIGURE A4.1 Share of people (15-64) working full-time for: Note: Survey questions reads: Number of months spent in full-time work during the previous calendar year. Source: SILC. GOOD JOBS IN TURKEY 117 FIGURE A4.2 Distribution of adult equivalent household income (nominal, in logs), 2005-2009 density nominal household income (logs) Source: SILC. 118 GOOD JOBS IN TURKEY FIGURE A4.3 Sources of household income, by region, 2007-2010 Source: SILC. GOOD JOBS IN TURKEY 119 TABLE A4.1 Trends in income, by household type, 2006-2009 Notes: Adult equivalent household income, 2006 prices. The minimum level of living standards corresponds to the income of the bottom decile in 2006, 2468 TL. 120 GOOD JOBS IN TURKEY TABLE A4.2 Determinants of living standards Notes: Marginal effects of probit regressions are shown. For material deprivation, survey years 2006-2010 have been pooled; for low-income rate, years 2007-2010 were considered. Reference categories: 15-25 years old, no schooling, spent less than 3 years in paid work, male, rural area, unpaid family worker. Regional dummies and year dummies are included. Significant coefficients (5 percent) are in bold