Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 8
2 Challenges in Measuring Labor Market Conditions Across Countries PRESENTATION Economists Martín Rama and Raquel Artecona clearly highlighted the difficulty they encountered in obtaining, evaluating, and organizing accurate, timely information on labor markets around the world. Summarizing their paper (Rama and Artecona, 2002), they noted that people hold strong views on the effects of globalization and the impacts of labor market regulations, but these views are generally based on limited evidence. Some data are available (from censuses, household surveys, etc.), they said, but these data are plagued with problems. Rama and Artecona’s goal was to gather existing data for developing countries, analyze it, select what was internationally comparable, and make it available to researchers in a ready-to-use format. Rama presented their results—a database of labor market information on 121 countries, classified into seven broad groups.1 The database includes information on most “larger” countries, as well as countries with relatively good data and those with which Rama and Artecona were personally familiar. Because the boundaries of many nations have changed over 1 The seven groups include the industrial or developed countries and six geographic regions defined by The World Bank: Sub-Saharan Africa; East Asia and the Pacific Islands; Eastern Europe and Central Asia; Latin America and the Caribbean; the Middle East and North Africa; and the South Asian region.
OCR for page 9
the past 50 years, the data were selected, and adjusted, to reflect national boundaries in 1999. Labor market information was averaged for 11 five-year time periods, from 1945–1949 to 1995–1999. Describing the database variables, Artecona said that they sought to include information on labor market outcomes, as well as to “provide, hopefully, some light on the debate about [whether] labor market interventions are good or bad.” The variables included 44 indicators in seven broad groups: (1) labor force; (2) employment and unemployment; (3) wages and productivity; (4) conditions of work and benefits; (5) trade unions and collective bargaining; (6) public sector employment; and (7) labor standards (see Table 2-1). Artecona said that when they were selecting indicators, they chose those with clear definitions. The database does not include information on the share of the informal sector in the total labor force because there was no clear definition of what constitutes informal employment. Artecona said they drew on a wide variety of cross-country and country-specific sources, including unpublished statistical information gathered during their research for the World Bank. Confronted with many different, and sometimes conflicting, data sources, Artecona and Rama considered three alternative strategies: to document every break, discrepancy, and change in definition; to recalibrate the data in order to make it fully consistent; or to exercise judgment in selecting data that appear comparable and to report these data without comments. Artecona argued that the first strategy is self-defeating because no one uses labor market data that are extensively documented to conduct research. The second strategy is not practical, she said, except for a few indicators, such as wages. This left the third strategy. Carrying out this strategy, Artecona and Rama first constructed a raw database, with all relevant labor market information from all sources. Second, they organized the data into 121 country files, including those variables from the “raw” database that were related to the 44 labor market indicators. When sources conflicted, they selected the “better” source, based on what they stated was their expert knowledge of countries and data sources. They also dropped variables, observations, and sources that were clearly out of line with other sources. Finally, they summarized the information in each country file in the form of five-year period averages, from 1945–1949 to 1995–1999. Rama noted that, although he and Artecona were still adding to the
OCR for page 10
TABLE 2-1 Variables in the Database of Labor Market Indicators Across Countries 1. Labor force LFTALL Total labor force, in thousands of persons. LFTAGR Labor force in agriculture, in thousands of persons. LFTIND Labor force in manufacturing, in thousands of persons. LFRALL Labor force participation rate, in percentage of the population aged 15 to 64. LFRMLE Male participation rate, in percentage of the male population aged 15 to 64. LFRFEM Female participation rate, in percentage of the female population aged 15 to 64. 2. Employment and unemployment EMTALL Total employment, in thousands of persons. EMTIND Employment in manufacturing activities, in thousands of persons. UNRALL Total unemployment, in percent of the labor force. UNRMLE Male unemployment, in percent of male labor force. UNRFEM Female unemployment, in percent of female labor force. UNTFST Unemployed seeking their first job, in thousands of persons. HRSWRK Hours of work per week, in nonagricultural activities. 3. Wages and productivity WGEAGR Wage per worker in agriculture, in current U.S. dollars per year. WGEIND Labor cost per worker in manufacturing, in current U.S. dollars per year. PRDIND Value added per worker in manufacturing, in current U.S. dollars per year. WGEGOV Wage of government employees, in current U.S. dollars per year. WGEMIN Minimum wage, in current U.S. dollars per year. 4. Conditions of work and benefits MATLVE Duration of maternity leave, in equivalent days with 100 percent earnings. ANNLVE Minimum annual leave with pay after one year of work, in working days. ACCDNT Number of workers injured at work, in thousands of persons per year. SSTYPE Mandatory coverage of the statutory pension scheme. SSCONT Social security contributions by employer and employee, in percent of salaries. SSCVGE Active contributors to old-age pension schemes, in percent of the labor force.
OCR for page 11
SSREVN Social security contributions by employers and workers, in percent of GDP. UNBRPL Initial unemployment benefit, in percent of earnings before job loss. UNBDUR Maximum duration of continuous unemployment benefits. SVCPAY Mandatory severance pay after three years of employment, in months of salary. 5. Trade unions and collective bargaining TUMMBR Total trade union membership, in percent of the total labor force. TUCVGE Coverage of collective bargaining agreements, in percent of salaried workers. STKNBR Number of strikes and lockouts per year. STKWRK Annual number of workers involved in strikes and lockouts, in thousands. STKHRS Annual work-days lost to strikes and lockouts, in thousands. 6. Public sector employment EMPCGT Employment in the central government, in thousands of persons. EMPGGT Employment in the general government, in thousands of persons. EMPPSR Employment in the public sector, in thousands of persons. 7. Labor standards ILOCNV Cumulative number of ILO conventions ratified by the country. CHLDLB Ratification of ILO Convention 138, on child labor. FORCLB Ratification of ILO Convention 29, on forced or compulsory labor. ABOLFL Ratification of ILO Convention 105, on the abolition of forced labor. EQLREM Ratification of ILO Convention 100, on equal remuneration. DISCRM Ratification of ILO Convention 111, on discrimination. ORGNZE Ratification of ILO Convention 87, on the right to organize. BRGAIN Ratification of ILO Cconvention 98, on the right to bargain collectively. database, it already had very good coverage for indicators on labor force, employment, and labor standards (see Table 2-2). For example, it included more than 700 observations for employment in agriculture and industry and labor force participation rates. On the other hand, coverage was weaker for indicators on trade union activities and conditions of work. For example, Rama said that one variable for which it was difficult to find data was severance pay after two years of uninterrupted private employment. Across world regions, the database provided good coverage for Latin America and South Asia, but “much worse” coverage of labor market information for Sub-Saharan Africa, the Middle East, and North Africa. Using
OCR for page 12
TABLE 2-2 Coverage by Variable and Region in the Draft Database of Labor Market Indicators Across Countries (as of July 2002) Region Countries LFTALL LFTAGR LFTIND LFRALL LFRMLE AFR 23 212 164 162 140 140 EAP 12 114 77 74 74 65 ECA 18 99 74 59 105 105 INL 23 223 175 153 153 138 LAC 21 196 147 144 126 126 MNA 19 174 131 126 102 102 SAS 5 49 35 35 30 31 ALL 121 1067 803 753 730 707 UNTFST HRSWRK WGEAGR WGEIND PRDIND AFR 23 10 17 15 97 87 EAP 12 13 26 13 71 59 ECA 18 13 25 51 66 34 INL 23 48 91 16 150 155 LAC 21 33 38 15 119 118 MNA 19 13 7 7 84 83 SAS 5 3 8 26 27 28 ALL 121 133 212 143 614 564 SSCONT SSCVGE SSREVN UNBRPL UNBDUR AFR 23 38 24 34 4 4 EAP 12 27 19 17 4 8 ECA 18 24 14 37 13 17 INL 23 46 9 107 49 56 LAC 21 56 58 81 8 20 MNA 19 43 23 25 7 14 SAS 5 9 8 0 1 1 ALL 121 243 155 301 86 120 EMPGGT EMPPSR ILOCNV CHLDLB FORCLB AFR 23 41 24 105 253 253 EAP 12 27 15 85 126 126 ECA 18 19 25 98 178 159 INL 23 42 26 252 253 253 LAC 21 35 17 150 231 231 MNA 19 26 17 106 208 208 SAS 5 10 8 46 55 55 ALL 121 200 132 842 1304 1285
OCR for page 13
LFRFEM EMTALL EMTIND UNRALL UNRMLE UNRFEM 140 15 31 16 5 5 74 51 59 55 31 31 105 59 58 28 16 16 138 187 183 155 193 157 126 53 56 78 61 61 105 36 21 31 14 14 30 21 18 25 18 18 718 422 426 388 338 302 WGEGOV WGEMIN MATLVE ANNLVE ACCDNT SSTYPE 37 41 52 28 30 34 22 26 22 10 35 19 20 33 32 14 21 18 33 102 69 43 56 46 25 64 75 29 50 42 17 30 37 18 23 29 13 5 14 9 6 8 167 301 301 151 221 196 SVCPAY TUMMBR TUCVGE STKNBR STKWRK STKHRS 13 67 7 30 28 30 3 56 5 40 33 31 3 43 4 18 17 18 0 181 50 63 66 88 12 75 10 46 46 39 17 42 0 11 11 11 22 23 0 27 27 27 70 487 76 235 228 244 ABOLFL EQLREM DISCRM ORGNZE BRGAIN 253 253 253 253 253 126 126 126 126 126 178 160 160 178 178 253 253 253 253 253 231 231 231 231 231 208 208 208 208 208 55 55 55 55 55 1304 1286 1286 1304 1304
OCR for page 14
Chile as an example, Rama explained how a country file is organized in the final database. Although analysis and comparison of variables in the larger “raw database” sometimes led easily to selection of the best source, this was not always the case. Fortunately, in Chile a study already existed in which researchers had tried to reconcile the figures. This study provided the information Rama and Artecona needed to select the best estimate of unemployment to include in the Chile country file in the final database. In conclusion, Rama noted that “only their use” will tell whether the data included in the database are helpful for research, and he added, “We want to report some encouraging results.” He cited several published and completed studies (Forteza and Rama, 2001; Rodrik, 1999; Freeman, 1994) that have drawn on preliminary versions of the data and have yielded meaningful results. Rama said his research with Forteza indicates that countries with relatively rigid labor markets adjust more slowly to economic reforms than those with more flexible labor markets. However, he said, “We found it is not minimum wages and mandatory benefits” that cause these rigidities, but rather the size of organized labor, which represents those who stand to lose as a result of economic reforms. DISCUSSION Responding to Rama and Artecona’s presentation, Yale economist T.N. Srinivasan commended “their valiant effort in putting together a data set.” However, he was not persuaded that the database would allow researchers to analyze the interaction between labor market policies and institutions on the one hand and economic growth, poverty, and inequality on the other hand. He was not convinced that the database would allow evaluation of the relationship between labor market conditions and openness to trade and foreign direct investment. Srinivasan said he was glad that Rama and Artecona’s data set has not yet fallen into the hands of cross-country regression analysts. He was unenthusiastic about such regressions because they are based on a common regression of data of varying quality from disparate economies. Therefore, he said, it would be inappropriate to make inferences about growth from such regressions unless they allow for both measurement errors and biases in some of the explanatory variables and for the possibility that there might be a two-way relationship between growth and some of the explanatory variables. For example, higher educational attainment of the labor force might lead to faster economic growth, while faster growth would enable the soci
OCR for page 15
ety to provide more education to its workers. Although in principle these econometric issues could be addressed, in practice it is either not possible to do so (for example, appropriate instruments are not available), or it is simply not done. Although he acknowledged that most countries have population censuses and household surveys, Srinivasan disagreed with the authors’ claim that they were able to derive employment and unemployment data from these sources. In India, he said, the census definition of “worker” has changed over time. It is difficult to adjust for such changes and for undercounts. As for data from establishment surveys, Srinivasan said it is “not a simple exercise” to put together a sample frame for drawing random samples. He said that economic censuses for the developing countries with which he is familiar do not survey all establishments and do not “provide a sensible base from which to put together a sample frame.” Srinivasan also noted problems with the approach of selecting the best source for each labor market variable in the database. First, for some countries, only one source may be available; by definition that source is the best, regardless of its accuracy. Second, choosing the best source “creates a selection bias” because more developed countries are likely to have more sources of data and data of better quality. As a result, selecting the best sources may affect in a nonrandom way the number of cells filled in the final data set. Srinivasan highlighted another problem: The data set does not include the informal sector, which accounts for the major share of employment in many developing countries. Without the informal sector, he said, “the data set is . . . not representative of the labor market conditions in developing countries.” Srinivasan also said that the strategy of not providing complete documentation of breaks, discrepancies, and changes of definition in the data was inappropriate. Economists who ignore problems such as changes in definition and breaks in the data “are not interested in the data, they are interested in an observation to throw into the regression.” Their deplorable practice should not be the reason for excluding footnotes and documentation and expecting other serious users to rely only on Rama and Artecona’s expert judgments. Srinivasan said it would be better to “hang it all out there and let the user decide what is appropriate.” Responding to Srinivasan’s comments, Rama explained why he and Artecona chose the approach they used. First, he said, he agreed with Srinivasan’s criticisms of cross-country regressions, and he said the database was not created to encourage economists to conduct more of these studies. Second, he agreed with Srinivasan that there are many problems with mea
OCR for page 16
surement error. However, he and Artecona found that the bias introduced as a result of measurement error in the data selected for each variable was not so large that the user would learn nothing from the variable. In addition, he said, there are so many different sources of measurement error, in so many different directions, that “our sense is that it can be treated as random error.” Although the database does not provide an indicator of the relative shares of employment in the formal and informal sectors for each country, Rama noted that some of the variables in the database do provide information on informal employment, such as labor force levels, labor force participation rates, unemployment rates, and work hours. In addition, one of the variables—the average wage of casual agricultural workers—refers exclusively to the informal sector. In further discussion, a committee member noted that the committee is charged with creating “an ongoing, living database.” He asked Rama what resources would be required to respond to the comments offered and to continue compiling and updating the database. Rama replied that a small team of experts could visit countries and arrange systematic meetings with labor lawyers, statistical officers, and others to identify and obtain the most accurate data sources. He suggested it might be appropriate to begin by visiting a core of about 30 countries. Another committee member described Rama’s response as “enormously optimistic,” because “a huge commitment of resources is usually necessary” to obtain reliable labor market data over time for just one country. Even an expert on a particular country may take a long time to fully understand such complex issues as the role of labor unions in that country’s economic growth. Developing an understanding of these complex issues internationally, in many different countries around the world, is “exceptionally difficult,” requiring much more time, funding, and expertise, he argued.
Representative terms from entire chapter: