## Guillermo Cruces, Gary S. Fields, David Jaume, and Mariana Viollaz

Print publication date: 2017

Print ISBN-13: 9780198801085

Published to Oxford Scholarship Online: June 2017

DOI: 10.1093/oso/9780198801085.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy).date: 18 October 2018

# Data and Methodology

Chapter:
(p.20) 2 Data and Methodology
Source:
Growth, Employment, and Poverty in Latin America
Publisher:
Oxford University Press
DOI:10.1093/oso/9780198801085.003.0002

# Abstract and Keywords

This study is based on microeconomic data from more than 150 household surveys, five million households, and eighteen million persons contained in the SEDLAC—Socio-Economic Database for Latin America and the Caribbean. These data cover the following sixteen Latin American countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Honduras, Mexico, Panama, Paraguay, Peru, Uruguay, and Venezuela. Based on these household surveys and the SEDLAC harmonization methodology, the study constructs comparable time series for a wide range of labour market, poverty, and income inequality indicators. It also employs aggregate macroeconomic indicators from two sources: the World Bank’s World Development Indicators and the United Nations Economic Commission for Latin America and the Caribbean’s database on social expenditure.

# 2.1 Data Sources

This study is based on microeconomic data from more than 150 household surveys, five million households, and eighteen million persons contained in the SEDLAC—Socio-Economic Database for Latin America and the Caribbean (CEDLAS and World Bank 2014). These data cover the following sixteen Latin American countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Honduras, Mexico, Panama, Paraguay, Peru, Uruguay, and Venezuela. Based on these household surveys and the SEDLAC harmonization methodology, we constructed comparable time series for a wide range of labour market, poverty, and income inequality indicators. In Chapters 3–4, we focus mainly on the changes from the initial to the final year in the period under study, listed for each country in Table 2.1. We present the indicators’ time series for each country in Appendix 1. For some countries, the period under study in this cross-country paper differs from the time period analysed in the corresponding country papers. The reason for using a different time period is the lack of comparability between the initial and final-year surveys. That was the case for Costa Rica, where we used 2000–9 as the period of analysis for all the labour market and income inequality indicators in this paper. For other countries, we used a different time period only for some particular indicators. Appendix 1 indicates with a vertical line when the country changed a classification so that it is not possible to use a consistent definition throughout the full time period.

Table 2.1 Latin American household surveys and period under study by country

Country

Isocode (two digits)

Initial year

Final year

Name of household survey

Argentina

AR

2000

2012

Encuesta Permanente de Hogares (2000–2)

Encuesta Permanente de Hogares-Continua (2003–12)

Bolivia

BO

2000

2012

Encuesta de Hogares—MECOVI

Brazil

BR

2001

2012

Pesquisa Nacional por Amostra de Domicilios

Chile

CL

2000

2011

Encuesta de Caracterización Socioeconómica Nacional

Colombia

CO

2002

2013

Encuesta Continua de Hogares (2000–5)

Gran Encuesta Integrada de Hogares (2008–13)

Costa Rica

CR

2001

2009

Encuesta de Hogares de Propósitos Múltiples

Dominican Republic

DO

2000

2012

Encuesta Nacional de Fuerza de Trabajo

EC

2003

2012

Encuesta de Empleo, Desempleo y Subempleo

Honduras

HN

2001

2012

Encuesta Permanente de Hogares de Propósitos Múltiples

Mexico

MX

2000

2012

Encuesta Nacional de Ingresos y Gastos de los Hogares

Panama

PA

2001

2012

Encuesta de Hogares

Peru

PE

2003

2012

Encuesta Nacional de Hogares

Paraguay

PY

2001

2013

Encuesta Permanente de Hogares (2002–13)

SV

2000

2012

Encuesta de Hogares de Propósitos Múltiples

Uruguay

UY

2000

2012

Encuesta Continua de Hogares

Venezuela

VE

2000

2012

Encuesta de Hogares Por Muestreo

Note: Venezuela’s surveys over 2000–6 are part of SEDLAC. From 2007 onwards, we carried out our own processing.

Source: SEDLAC (CEDLAS and World Bank 2014).

In this book, we also employ aggregate macroeconomic indicators from two sources: the World Bank’s World Development Indicators (World Bank 2014) and the United Nations Economic Commission for Latin America and the Caribbean’s database on social expenditure (UN-ECLAC 2015).

# (p.21) 2.2 Labour Market Indicators

The main purposes of the analysis are to determine whether each labour market indicator has improved or deteriorated over time on a country-by-country and cross-country basis, and what are the determinants and correlates of these changes. We use, in total, sixteen labour market indicators which we assign to one of two different categories: employment and earnings indicators, and poverty and income inequality indicators. For the employment and earnings indicators, we judge a welfare improvement to have taken place if we find:

Unemployment:

• A decrease in the unemployment rate.

Occupational composition:

• A decrease in the share of low-earnings occupations.

• An increase in the share of high-earnings occupations.1

• (p.22) An increase in the share of wage/salaried employees.

• A decrease in the share of self-employment.

• A decrease in the share of unpaid family workers.2

• A decrease in the share of workers in low-earnings sectors.

• An increase in the share of workers in high-earnings sectors.3

• A decrease in the share of workers with low levels of education.

• An increase in the share of workers with high levels of education.4

• An increase in the share of workers registered with the social security system.

Labour earnings:

• An increase in mean labour earnings.

For the poverty and income inequality indicators, we judge a welfare improvement to have taken place if we find:

Poverty and inequality:

• A decrease in the 4 dollars-a-day poverty rate.

• A decrease in the 2.5 dollars-a-day poverty rate.

• A decrease in Gini coefficient of household per capita income.

• A decrease in Gini coefficient of labour income.

More specifically, these indicators are defined as follows.

The unemployment indicator is defined following the ILO guidelines: it represents the share of unemployed persons divided by the number of persons in the labour force. A person is unemployed if s/he is 15 years old or more and during the reference period (usually one month, but it depends on the survey of each country), s/he was without work, available for work, and seeking work. A fall in the unemployment rate is classified as an improvement in the labour market.

Occupational groups are defined by means of a two-step process. First, for each country, we identify the following occupational categories:5 management; professionals; technicians and associate professionals; clerical; service and sales workers; agricultural, forestry, and fishery workers; craft and related trades workers; plant and machine operators and assemblers; elementary and armed forces. Second, we classify them into low-earnings, medium-earnings, and high-earnings occupations. For each country, the low-earnings occupations are defined as the three occupational categories with the lowest mean earnings (p.23) during the analysed period, the high-earnings occupational categories are the three occupations with the highest mean incomes, and the rest are classified as medium-earnings occupations. A fall in the share of low-earnings occupations and an increase in the share of high-earnings occupations imply an improvement in the labour market.

Occupational position is classified into four categories: employer, wage/salaried employee, self-employed, and unpaid worker. Given the nature of labour markets in Latin America, the analysis of the employment structure according to occupational positions identifies as improvements in the labour market the following situations: a decrease of self-employment, a decrease in the share of unpaid family workers, and an increase in the share of wage/salaried employees.

Sector of employment is also classified by means of a two-step procedure. We first identify ten sectors: primary activities; low-tech industry; high-tech industry;6 construction; commerce; utilities and transportation; skilled services; public administration; education and health; and domestic workers. We further classify the sectors according to the shares of workers in low-, medium-, and high-earnings sectors, using the same criteria as in the case of the occupational groups. An increase in the share of high-earnings sectors and a decrease in the share of low-earnings sectors represent improvements in the labour market in our analysis.

With respect to the educational level of employed workers, we define three categories for the analysis: low (eight years of schooling or less); medium (from nine to thirteen years of schooling); and high (more than thirteen years of schooling). An increase in the education of the employed population is considered an improvement in the labour market, as the share of workers that are expected to receive high levels of earnings increases and the share of workers with low earnings levels decreases.

We also classify the employed population according to whether they are registered with the social security system or not. In some of the countries, only wage and salaried employees are asked about registration in the social security system. We assume that it is better for employed workers to be registered, and thus an increase in this indicator is classified as an improvement in the labour market.

Labour earnings are expressed on a monthly basis in 2005 purchasing power parity (PPP) dollars. Higher earnings represent an improvement in the labour market.

Poverty and inequality are calculated as follows. Poverty rates are based on the international poverty lines of 4 dollars a day and 2.5 dollars a day (all in (p.24) PPP dollars), and represent the poverty and extreme poverty levels respectively, often used in Latin America. These poverty indicators are based on household income per capita. Household income is the sum of labour income plus non-labour income, which includes capital income, pensions, public and private transfers, and the imputed rent from own-housing. Income inequality is calculated using the Gini coefficient of household per capita income and of labour earnings among employed workers. Some caveats should be stated regarding the use of household surveys to calculate inequality indices as the Gini coefficient. Household surveys suffer from non-response to income questions, underreporting of incomes, and lack of coverage of very high incomes. All these problems may have impacts on calculated inequality levels and trends. In this book we use income variables from SEDLAC databases which are based on ‘raw data’ from household surveys, without applying any adjustment or correction procedure.

To sum up, changes in labour market indicators in Latin American countries during the 2000s are evaluated using the following criteria. Improvements in labour market conditions are associated with: a decrease in unemployment; increases in the shares of high-earnings occupations, wage/salaried employees, workers in high-earnings sectors, and workers with high levels of education; an increase in monthly labour earnings; declines in the shares of low-paid occupations, unpaid family workers, self-employed, low-earnings sectors, and workers with low levels of education; and declines in poverty rates and inequality indicators. Worsenings in labour market conditions are associated with changes in labour indicators in the opposite direction.

# 2.3 Macroeconomic Indicators

We also use data on macroeconomic variables to correlate them to the changes in labour market indicators described in section 2.2. These data come from two sources. First, from the World Bank’s World Development Indicators (WDI), we use: GDP per capita in the initial year; agriculture as a percentage of GDP; industry as percentage of GDP; services as a percentage of GDP; final consumption expenditure as a percentage of GDP; exports as a percentage of GDP; terms of trade; foreign direct investment as a percentage of GDP; and revenues from natural resources as a percentage of GDP. Second, from the United Nations Economic Commission for Latin America and the Caribbean (UN-ECLAC 2015) database on social expenditure, we use: expenditure in education and health as a percentage of GDP; public expenditure in social security as a percentage of GDP; and stock of public debt as a percentage of GDP. For all macroeconomic variables with the exception of GDP per capita in the initial year, we use data on the initial and final years and calculate the annualized change.

# (p.25) 2.4 Variables and Notations

We denote each of the K labour market indicators as Yk and each of the j macroeconomic variables as Xj. In the following analysis, we will use this notation:

Xijt: Macroeconomic variable k for country i at time t.

Yikt: Labour market indicator k for country i at time t.

%Δ‎Xij: Annualized percentage change of macroeconomic variable j for country i from initial to final year.

Δ‎Xij: Annualized change in percentage points of macroeconomic variable j for country i from initial to final year.

%Δ‎Yik: Annualized percentage change of labour market indicator k for country i from initial to final year.

Δ‎Yik: Annualized change in percentage points in labour market indicator k for country i from initial to final year.

Zi : Percentage of labour market indicators that improved for country i from initial to final year.

Note that the operator %Δ‎ embodies an annualized percentage change. We calculate annualized percentage changes for GDP per capita, labour earnings, Gini coefficients, and terms of trade. For the rest of the indicators, the operator Δ‎ is used, indicating annualized changes in percentage points. For example, annualized changes in percentage points include the change in unemployment, in the share of workers registered with the social security system, or in industry’s share of GDP.

We calculate these changes as follows. Let initial year be t0 and final year be t1. Then:

$Display mathematics$
(1)

(p.26) As a way to summarize the evolution of the large number of indicators covered in each country study, we devised a measure Zi based on the percentage of the available labour market indicators for each country over the period under study which exhibited a statistically significant improvement at the 5 per cent level.7 We express Zi as a percentage instead of the actual number of indicators that increased because not all indicators are available for all countries in every year. This measure provides a general direction of change in the labour market. The costs of this simple synthetic index are that it implicitly assigns an equal weight to each indicator, and it does not take into account the magnitude of the changes (only if the change was statistically significant or not). Nonetheless, this index provides a handy summary indicator of labour market improvements in each country, and so we make extensive use of it in the analysis that follows.

# 2.5 A Note on Causality versus Correlation

The change in a macroeconomic variable j (Δ‎Xj or %Δ‎Xj) and the change in a labour market indicator k (Δ‎Yk or %Δ‎Yk) may be associated with each other either because Δ‎Xj causes Δ‎Yk or because the two of them are caused by a third factor. An example of Δ‎Xj causing Δ‎Yk would be a situation in which a shock in terms of trade brings about an increase in the demand for labour and in mean labour earnings. An example of Δ‎Xj and Δ‎Yk being caused by a third factor would be a situation in which training more workers in occupations where shortages exist results in higher exports and an improvement in employment composition in favour of high-earnings occupations.

We implicitly assume throughout the analysis that there is not reverse causation: that is, that changes in labour market indicators do not affect macroeconomic variables (or at least not directly). It is a judgement call whether to make causal interpretations or to be more cautious and choose wording in terms of correlations between variables, and we have done some of each.

References

Bibliography references:

CEDLAS and World Bank (2014). SEDLAC—Socio-Economic Database for Latin America and the Caribbean. Centro de Estudios Distributivos, Laborales y Sociales, Facultad de Ciencias Económicas, Universidad Nacional de La Plata and World Bank Poverty Group LCR. Available at <http://sedlac.econo.unlp.edu.ar/eng/index.php>, accessed 2014.

(p.27) UN-ECLAC (2015). CEPALSTAT, United Nations, Economic Commission for Latin America and the Caribbean. Available at <http://estadisticas.cepal.org/cepalstat/WEB_CEPALSTAT/Portada.asp?idioma=i>, accessed April 2015.

World Bank (2014). World Development Indicators. Available at <http://data.worldbank.org/data-catalog/world-development-indicators>, accessed April 2014. (p.28)

## Notes:

(1) The residual category is the share of medium-earning occupations.

(2) The residual category is the share of employers.

(3) The residual category is the share of medium-earning sectors.

(4) The residual category is the share of medium-educated workers.

(5) This is the International Standard Classification of Occupations of 2008 (ISCO-08) at a one-digit level. In the case of Argentina, this classification cannot be obtained from household survey data. Argentina is then excluded from the analysis of changes in the occupational composition of the employed population.

(6) For Bolivia and Paraguay, we cannot distinguish between low- and high-tech industries.

(7) The significance of changes is computed as a mean difference test between the initial and the final year for each country in the sample.