Unemployment in Latin America and the Caribbean

Abstract

This study constructs a new data set on unemployment rates in Latin America and the Caribbean and then explores the determinants of unemployment. We compare different countries, finding that unemployment is influenced by the size of the rural population and that the effects of government regulations are generally weak. We also examine large, persistent increases in unemployment over time, finding that they are caused by contractions in aggregate demand. These demand contractions result from either disinflationary monetary policy or the defense of an exchange-rate peg in the face of capital flight. Our evidence supports hysteresis theories in which short-run changes in unemployment influence the natural rate.

Data Appendix

1.1 Cross-country Comparable Unemployment Rates from the IADB

The unemployment data from the Interamerican Development Bank are based on household surveys in each country in its sample. The IADB seeks to produce harmonized statistics that are as comparable as possible across countries and over time. The data are described in IADB (2004).

Unemployment Definition: The IADB measures unemployment as the number of unemployed divided by the labor force (employed plus unemployed). For most countries, an individual’s status is determined by his situation during a “reference week.” He is counted as unemployed if he was not employed and did not search actively for work during the reference week. Active search includes activities such as contacting potential employers or employment agencies, interviewing for jobs, or filling out applications. The labor force is restricted to ages 15 through 64.

In Colombia and Mexico, the reference period is 1 month rather than 1 week; in Chile, it is 2 months. A longer reference period increases the unemployment rate. Brazil’s monthly employment survey reports unemployment rates based on one-week and one-month reference periods in June and July, 2002; the unemployment rate based on 1 week is 0.92 times the rate based on 1 month. Based on this example, we multiply all unemployment rates in Colombia, Mexico, and Chile by 0.92 to make them comparable with other countries. IADB staff have told us that this adjustment is reasonable.

Our adjusted version of the IADB data is presented in Table 6.

Table 6 IADB Unemployment rates

1.2 New Data Set of Unemployment Rates

To derive our data, we have gone country by country to figure out how unemployment was measured in different periods. We have made judgments about which changes in methodology are small enough to ignore, and how to adjust for larger changes. In some cases we can splice different unemployment series together using periods in which they overlap. When in doubt, we have sought advice from people at the agencies that produce unemployment data. We have pieced together data from country-specific sources –central banks, labor ministries and national statistical agencies– and international agencies such as the International Labor Organization. Where a methodological change appears significant but we cannot reliably estimate its effects, we discard the shortest portion of the data that yields a consistent series

Here we report the sources of data, the definition of unemployment, and our adjustments to the series for each country. Table 7 reports the data.

Table 7 New dataset of unemployment rates

Argentina

Sources: ILO (International Labor Organization) and INDEC (National Institute of Statistics and Census). Surveys cover Gran Buenos Aires and include people ages 10+. Definition: No job and searched actively during the reference week. Prior to 2003, unemployment rates are averages based on surveys in May and October. In 2003, several methodological changes were introduced: the frequency of surveys was increased to one per quarter; some types of female labor that had previously been ignored were included in employment; and the definition of job search of the unemployed was broadened. At one point in time, the INDEC reports results for both versions of the surveys: it reports the new series for the second quarter of 2003 and the old series for May 2003. The ratio of the unemployment rates in the new and old data is 1.14. Therefore, to correct for the break in the series in 03, for 03–07, we average outcomes for the second and fourth quarter and divide the figure by 1.14.

Bolivia

Sources: ILO and INE (National Institute of Statistics). The age covered by the surveys is 10+, while the geographical coverage is “main towns” prior to 1996 and urban areas thereafter. After analyzing the data, we decided that the change in geographical coverage was small enough to ignore. Definition: No job and searched actively during the reference week. The unemployment rate for 1998 is missing; in Table 7, we impute this number by averaging the unemployment rates in 1997 and 1999.

Brazil

Source: IBGE (Brazilian Institute of Geography and Statistics), PME (Monthly Employment Survey). The surveys cover people 15+. Geographical coverage: Metropolitan regions of Recife, Salvador, Belo Horizonte, Rio de Janeiro, São Paulo and Porto Alegre. The figures we report are averages of monthly rates. Definition before 2003: No job and searched actively during the reference week. In 2003, a number of methodological changes were introduced, including new definitions of the geographical areas covered by the survey and an extension of the reference period for search to 30 days. The IBGE performed both versions of the survey from March through December 2002; the average ratio of unemployment rates is 1.47. Therefore, we divide the unemployment rates for 2003 and later by 1.47 to make them comparable to the older data.

Chile

Sources: Encuesta de ocupación y desocupación, Universidad de Chile. The survey covers Gran Santiago and is performed in June of each year. Ages 14+. Definition: No job and searched actively during the reference week. There are no breaks in the original series.

Colombia

Sources: ILO and DANE (Administrative National Department of Statistics). We use the survey from September of each year. Ages 12+. Geographical coverage includes seven cities: Bogota, Medellin, Cali, Barranquilla, Bucaramanga, Pasto and Manizales. The weights on the cities changed in 1991; we ignore this change. Definition before 2001: No job and searched actively during the last year (thus, includes “hidden unemployment” or “discouraged workers” who would not be counted as unemployed in many countries). In 2001, there were several methodological changes including a redefinition of unemployment to require search during a reference week. Both versions of the unemployment rate were reported in 2000; the ratio of the old and new rates is 1.19. Therefore, we multiply unemployment rates in 2001 and later by 1.19. In 2006, new methodological changes were made; nevertheless, CEI (an independent commission of experts 2008) concluded that the impact of the changes on the urban unemployment rate is statistically insignificant. Thus, we ignore the break.

Costa Rica

Sources: ILO and INEC (National Institute of Statistics and Census). National survey in July of each year. Ages 12+. Definition before 1987: No job and looked for one during the reference week or did not look for a job during the reference week for “circumstantial reasons” but looked for one “in the past”. Starting in 1987, “In the past” is restricted to 5 weeks. This methodological change is similar to the narrowing of the unemployment definition in Colombia and Panama. Based on these countries, we estimate that the old definition raises unemployment rates by a factor of 1.1. Therefore, we divide the pre-87 data by 1.1.

Dominican Republic

Source: Labor Force National Survey, Central Bank of the Dominican Republic. National rates, ages 10+. Definition: Includes hidden unemployment, that is, no job search required to be counted as unemployed.

Ecuador

Source: ILO. Survey in November (except July in 1993 and 2001, December in 2007). Ages: 12+ until 1990, 10+ thereafter; we ignore this change. Geographic coverage is urban areas of Quito, Guayaquil and Cuenca. Definition: No job and active search during a reference period of 5 weeks.

El Salvador

Source: ILO. In the original data, unemployment rates are national except for 1988–1992, when they are urban. IADB data suggest that urban rates are 1.1 times national rates. Therefore, we divide the unemployment rates for 1988–92 by 1.1. Ages 10+. Definition: no job and have actively searched for a job (reference period is not explicit).

Guatemala

Source: ILO. National rates, ages 10+. Definition: No job during reference week and searched actively during the 4 preceding weeks.

Honduras

Source: INE (National Institute of Statistics). Survey of urban areas in September (except March in 1993 and 2001). Ages 10+. Definition: No job in the reference week and searched actively during the reference week. Unemployment rate for 2000 is missing; in Table 7 we impute this number by averaging the unemployment rates in 1999 and 2001.

Jamaica

Sources: ILO. National rates. Definition: “The unemployed comprise all those aged 14 years and over who were looking for work, wanting work and available for work. Persons looking for work must have made a positive attempt to seek a job such as: registration at employment agency; visiting job sites in search of a job; applying in person to prospective employers; putting advertisements in any public press or place; writing letters of application; asking someone to try to find a job; making investigations with a view of starting own farm or business.”

Mexico

Sources: ECSO (Continuous Survey on Occupation), 1973–1984, and ENEU (National Employment Urban Survey), 1985–2004, ENOE (National Occupation and Employment Survey) 2005–2007. The ECSO data cover three cities: Mexico City, Guadalajara and Monterrey. ENEU provides unemployment rates for individual cities; we construct a weighted average for the three cities using population estimates from the 1990 census. Ages 12+. Definition: No job and searched actively during the reference week. Before 1985, the unemployed include persons waiting to start a job in 30 days; persons who expect to return to a previous job within 30 days; and unpaid family workers working less than 15 h per week. Starting in 1985, the first two groups are counted as employed and the third is out of the labor force. Both versions of the unemployment rate are reported for 1984; the ratio of the old definition to the new one is 1.05. Therefore, we divide the data before 1985 by 1.05.

Starting in 2005, the information comes from a new survey, the ENOE. With this survey, Mexico started producing OECD comparable labor market indicators. There are several changes. We highlight two, but there are more (for a summary see “ENOE: ¿Cómo se hace la ENOE? Métodos y Procedimientos”, Inegui, 2007). For instance, in the previous surveys persons declaring that they were currently without a job but about to enter a new job (in less than a month) were declared employed. In the ENOE, following OECD’s practices, they are coded as unemployed. Another example: in the previous surveys persons without a formal job but declaring that they run an independent business or work, were declared employed. In the new survey, they are declared unemployed if in the previous week they did actually not work in their independent activity and received no income from that source. Using information for dates where the two surveys overlapped, the INEGI (National Institute of Geography and Statistics) estimated that the new method produces figures at the national level that are 49 % larger. Thus, we divide the figures of the ENOE (again a weighted average of the three main cities) for 2005–07 by 1.49 to make them comparable to the rest of the series.

Panama

Sources: UN Yearbook of Labour Statistics (1963–69), ILO (1970–99), Contraloría de Panamá (2000-). National rates, ages 15+. Definition: includes hidden unemployment. The unemployment rate is missing for 1980–81 and 1990. In Table 7, we use the 1979 rate for 1980–1981. (We did not use the 1982 rate because of the likely effect of the Latin American debt crisis.) We impute a rate for 1990 by averaging the rates for 1989 and 1991.

Paraguay

Sources: ILO (1979–96), LAC Statistical Yearbook (1997-). Geographical coverage is Asunción Metro area until 1994 and urban areas thereafter. Ages 12+ for 1979–1992 and 1994; otherwise, 10+. We ignore these changes. Definition: No job in reference week and searched actively during reference week. The unemployment rate for 1981 is missing; we use the average of 1980 and 1982.

Peru

Sources: MTPS (Ministry of Labor and Social Promotion), 1970–94; INEI (National Institute of Statistics), 1995-. Unemployment rates for Lima, ages 14+. Definition: No job in reference week and searched actively during reference week. Missing data for 1978, 1985, and 1988 are imputed by averaging adjacent years.

Trinidad and Tobago

Sources: Yearbook of Labour Statistics and ILO. National survey, ages 15+. Definition : includes hidden unemployment. (We have discarded data before 1987, which do not count first-time job seekers as unemployed, because we do not know how large an adjustment to make to these data.)

Uruguay

Sources: Yearbook of Labour Statistics and INE (National Institute of Statistics). Unemployment rates for Montevideo, ages 14+. Definition: No job during reference week and searched actively during reference week. Starting in 1981, domestic workers are not counted as employed; this change appears small enough to ignore. The unemployment rate in 1975 is missing; we impute it by averaging 1974 and 1976.

Venezuela

Sources: ILO and INE (National Institute of Statistics). National rates, ages 15+. We calculate annual unemployment rates as averages of rates for the first and second halves of each year. Definition: Includes hidden unemployment.