Labor Income Profiles Sang-Hyop Lee November 5, 2007 Prepared for NTA 5 th Workshop SKKU, Seoul, Korea
Outline of Panel Discussion I.Development of New Methodology (self-employment income) II.Analysis 1.Cross-Section Comparison 2.Time-Series Analysis (X) III. More In-Depth Analysis (from Ogawa) (X) IV. Remittances (from Salas) V. Other Issues (smoothing, etc)
Issues in estimating self-employment income –Labor markets in low-income countries (Rosenzweig 1988) Large proportion of agricultural sector Low proportion of wage earners and large proportion of family enterprises or unpaid family workers Empirical issues; especially estimating labor income for unpaid family workers I. Development of New Methodology (self-employment income)
Unpaid Family Workers Old Method –Don’t impute. –It may underestimate/overestimate the share of earnings for age x New Method –Estimate using the age profile of earnings of employees as a share to allocate household self-employed income to self- employed workers including unpaid family workers. Ex) A household (2/3 of household self-employed income = 30) AgeEarnings per employeeImputed 18 (unpaid) (self emp.)40020
Philippines (2002) Indonesia (1996) Thailand (1996) Taiwan (2001) Imputed?YesNoYesNoYesNoYesNo Mean age Share of life time earnings Under Over Under
II. Analysis-Comparative Summary statistics for 18 economies Age earnings profile for 20 economies –Suggestions for outliers (explain, estimate another year, etc) Wages vs. self-employment income
Labor IncomeMeanPeakMedian25th75thInterqrtile % 65 Austria Brazil Chile China Costa Rica France India Indonesia Japan Mexico Philippines S.Korea Slovenia Sweden Taiwan Thailand Uruguay US Average
Outliers?
Thick Flat Tails
Cliffhanging (at a certain old age)
Start late, exit late
Steep in early ages
The winner and the runner-up
Why do they differ? Mechanical decomposition 1.(Y/N) =(Y/E) * (E/N) 2.Per capita labor income = Earnings per employee * (effective) labor force participation rate 3.(Y/N) =w*(Y/N)employee+(1-w)(Y/N)self-employed Thus per capita labor income profile depends on –Share of self-employed in the economy –Composition: Labor force participation rates (LFPRs) by age (inverse U), working hours by age (inverse U), unemployment rate by age –Productivity: Age specific productivity (concave/inverse U) (health, technological change, OJT), selection effect (hazard rate may increase over time) –Institution (minimum wage, seniority-based wage system) Decisions made by three demographic groups (women, children, and elderly) are most important
Relationship with Macro Variables Level of development (per capita GDP) Share of sector (e.g. agricultural sector, service sector, etc) Enrollment of secondary schooling Old age dependency Pension / Tax enforcement (not done)
II-2. Time Series Analysis Has an advantage –Consistent data sets & definitions –Decomposition across years –Policy change analysis
Source: provided by Ron Lee
Decomposition of the Change in Per-Capita Labor Income, Chile,
Source: provided by Ron Lee
Summary The share of self-employed income is an important factor affecting profiles for developing countries. Decisions made by women, children, and elderly might be important in shaping the labor income profiles across countries and over time. These decisions may be somewhat related with the level of development, but there are other factors affecting the relationship.
III. More In-Depth Analysis IV. Remittances Age earnings profile also reflects a host of vital economic and social conditions. Regular vs. Non-regular or Part-time vs. Full-time distinction (share of full-time, regular workers decrease in Japan) Demand side or macro economic condition (lack of job opportunities) Women’s labor force participation Other sectoral allocation of the labor force Age profile of compensation from/to ROW may be also different from those of residents.
V. Other issues Smoothing –Use SUPSMU in the R statistical package. Smoothing spans are determined on an ad hoc basis. –Any ages with a profile value of zero are left out of the calculation and added to the series after smoothing. For example when a survey only covers ages 14 and above, all values below 14 were set identically to zero.
Remaining Issues Refining estimation Other analysis –How does labor income interact with private consumption and private transfer? –How policy matters? Public pension programs Education (e.g. mandatory schooling) –How does labor income profile differ by education/gender/place of residence/living arrangement?