Local job multipliers in Turkey Altan Aldan William Maloney Josefina Posadas Temel Taskin (Central Bank of Turkey) (World Bank) (World Bank) (World Bank) The views expressed here belong to.
Download ReportTranscript Local job multipliers in Turkey Altan Aldan William Maloney Josefina Posadas Temel Taskin (Central Bank of Turkey) (World Bank) (World Bank) (World Bank) The views expressed here belong to.
Local job multipliers in Turkey Altan Aldan William Maloney Josefina Posadas Temel Taskin (Central Bank of Turkey) (World Bank) (World Bank) (World Bank) The views expressed here belong to the authors and should not be interpreted as the official views of the World Bank or Central Bank of Turkey. Posadas, World Bank, 1818 H Street NW, Washington DC, USA; [1] Corresponding author: Josefina Tel: +1 202 458 7279; [email protected] Motivation • Job multiplier: Every time a local economy generates a new job by attracting a new business/growing firms in the tradable sector, a significant number of additional jobs are created in the non-tradable (service) sector. • Job multiplier is larger for high productivity jobs: high levels of human capital, high-tech industries. • New non-tradable jobs created in local labor markets with agglomerations tend to have higher wages than similar non-tradable jobs in local labor markets without agglomeration economies. • Thus, new high-skilled jobs benefit the whole population of the local labor market (through multiplier effect) and not only those in highskilled jobs. • Governments devote significant effort and resources to attract new businesses and job creation. • Estimating multiplier effects is important to use resources efficiently and to understand effects of these policies. 2 Turkish context (1) Important regional variation in the growth of tradable and non-tradable employment tradable 1990-2000 non-tradable 1990-2000 3 Turkish context (2) Important regional variation over time in the ratio of change in nontradable to tradable employment Ratio NT/T 1990-2000 Ratio NT/T 2000-2011 4 This paper • Measures job multiplier effect in Turkey and compares it to USA and Sweden • Based on Moretti and Thulin (2013) methodology • Results are divided in 2 parts due to data limitations • 1985-2000 • 1985-2011 5 Methodology (1) • Estimation problem: • Job creation in the tradable and non-tradable sector responds to common unobserved factors cofounding effects. • Might be reverse causality from non-tradable sector to tradable sector under full-employment and no labor mobility across labor markets • Solution: Instrument: predicted employment growth at the level of the district and industry based on • Industry distribution in the local labor market at the beginning of the period, • National industry growth that isolates the district-specific effect. 6 Methodology (2) Econometric model: ∆𝐸𝑐,𝑡 𝑁𝑇 = 𝛼 + 𝛽∆𝐸𝑐,𝑡 𝑇 + 𝛾𝑑𝑡 + 𝜀𝑐,𝑡 where • • • ∆𝐸𝑐,𝑡 𝑁𝑇 is change in employment in the non-tradable sector in city c and time t and t-s ∆𝐸𝑐,𝑡 𝑇 is change in employment in the tradable sector in city c and time t and t-s 𝑑𝑡 is a dummy variable for the second period Instrument: ∆𝐸𝑐,𝑡 𝑇 = 𝑇 𝑇 𝑗 𝐸𝑐,𝑗,𝑡−𝑠 ∆ln(𝐸𝑗,𝑡 − 𝐸𝑐,𝑗,𝑡 𝑇 ) where • • 𝑇 𝐸𝑐,𝑗,𝑡−𝑠 is the number of jobs in industry j in city c at time t-s ∆ln(𝐸𝑗,𝑡 𝑇 − 𝐸𝑐,𝑗,𝑡 𝑇 ) is the change in national (excluding city c) employment in industry j between year t and t-s. 7 Data (1) • Ideal data: • Long time series to capture the structure of the economy (as opposed to the cyclical variation) 20 years, • Good regional disaggregation to define local markets precisely NUTS3 or lower, • Good sector/industry disaggregation NACE3, • Employment by education level of education, • Type of employment wage vs. self-employment, • Wages (by the categories of above), • Public expenditure in industry promotion. • Population census: • At least 3 periods and a long time coverage (1985;1990;2000; 2011), • Good regional disaggregation (district), • Poor sector disaggregation (only 8 tradable sectors), • Good education and employment type. 8 Variable definitions Based on Census data: • Jobs: any person that reports being employed • Counts formal and informal; full-time and part-time; wage work and self-employment, • Skilled job: if the worker has secondary or higher level of education, • Industry classification: • Excludes agriculture, mining, military and government, • Includes 8 tradable sectors, • Includes 16 non-tradable sectors. • Geographic unit: • District including metropolitan areas, • Small districts (less than 5000 workers) are merged with the closest large district. 9 Results (1) Population Census Data: 1985, 1990, 2000 Effect of tradable on non-tradable (T-NT) Effect of tradable on tradable (T-T) 0.27 3.01 1.97 OLS IV (Excluding Istanbul) 2.07 0.18 IV(Including Istanbul) OLS Each tradable job generates 3 non-tradable jobs in Turkey. 0.19 IV (Excluding Istanbul) IV(Including Istanbul) Each tradable job generates 0.27 tradable jobs in Turkey. 10 Results (2) Population Census Data: 1985, 1990, 2000 Effect of tradable on non-tradable Effect of tradable on tradable 0.36 3.90 0.32 3.25 0.19 0.87 0.75 1.08 0.06 -0.18 Total Non-tradable Skilled Non-tradable Jobs Jobs Skilled Tradable Jobs 0.18 -0.03 Unskilled Nontradable Jobs Unskilled Tradable Jobs Each skilled tradable job generates 3.90 non-tradable jobs. Total Tradable Jobs Skilled Tradable Jobs Skilled Tradable Jobs Unskilled Tradable Jobs Unskilled Tradable Jobs Each skilled tradable job generates 0.32 tradable jobs, Of which, 3.25 are skilled non-tradable jobs, And each unskilled tradable job generates 0.18 tradable jobs. and 0.75 are non-skilled tradable jobs. 11 Results (3) Population Census Data: 1985, 1990, 2000, 2011 3.1 Drop due to aggregation Constant over time* 2.81 2.74 district province 1985-2000 province 1990-2011 *No statistically significant variation of the job multiplier over time 12 Comparison Multipliers USA Sweden Turkey Average, tradable on non-tradable 1.6 0.5 3.01 High Skill, tradable on non-tradable 2.5 2.8 3.90 High Tech, tradable on non-tradable 4.9 1.1 - Average, tradable on tradable 0.26 0.33 0.18 Potential factors limiting the multiplier effects of Sweden: • • • Lower labor mobility, Lower wage differentials between high- and low-skilled workers, Generous social assistance and unemployment insurance. Potential factors limiting the multiplier effects of Turkey: • • Lower labor mobility for high-skilled workers, maybe due to amenities in large cities? Large hiring and firing costs, maybe mitigated through sizable informal sector. Potential factors strengthening the multiplier effects of Turkey: • • High income inequality and skill premium, Low social expenditures relative to GDP. Notes: Turkey: results based on censuses 1985-2000; USA: results from Moretti (2010); Sweden: results from Moretti and Thulin(2013) 13 Job multiplier for evidence-based policy Use the multiplier effect for policy-making • Tax incentives for new businesses, especially if geographically concentrated; • Design stimulus or counter-cyclical policies; • Promotion of innovation and high-value added industries; • Understand future demand of jobs and infrastructure resulting from industry growth in specific geographic areas; • Understand future internal migration as some districts grow and others fizzle. 14