Minimum Wages

We report nominal minimum wages in national currency, as a ratio to the mean wage, and, in some cases, relative to the median wage. All wages are reported on a monthly basis. The main data sources are IMF, OECD, Eurostat, ECLAC, Inter-American Development Bank, CIS statistics, Asian Development Bank, African Development Bank, and national statistics offices. Data on average wages were primarily collected from ILO KILM and ILO Laborsta, although for a range of countries also from national sources. Appendix II contains all sources of information on a country-by-country basis.

When minimum wages are set on other than a monthly basis, we convert them to monthly wages based on a number of assumptions (see Appendix I). These assumptions, such as a 40-hour working week (set by the ILO C47 Forty-Hour Week Convention, 1935) may not be fully appropriate in many developing countries, either because the C47 Convention has not been ratified, or because it is not applied. Thus, users are invited to adjust the monthly measures to the actual hours worked wherever additional information is available.

Several shortcomings in the measurement of minimum wages should be emphasized. First, they reflect only the formal sector, an important caveat especially for studies focusing on developing countries, where informal sectors can be large. Second, in a number of countries, there can be several minimum wages, differentiated across regions (such as in Indonesia), sectors (Sri Lanka), types of skill (Nepal) or type of enterprise (Vietnam). In these cases, we report the simple average of existing minimum wages.² In other countries, especially those with periods of high inflation (such as Belarus), minimum wages were reset several times during a given year, and we report only the ones in effect on the first of July 1 of the corresponding year. And third, even though many countries, including several in Europe, do not have statutory minimum wages, collective wage agreements often form de facto wage floors, so reporting minimum wage as zero would be misleading. In the database, we have marked explicitly where collective wage agreements are in place; however, further data collection in such countries would be fruitful.³

Minimum wages are also reported as ratios to average and, in a subset of countries, to median wages. Relating minimum wages to some measure of the aggregate level of wages is important for cross-country comparisons, but neither measure is without limitations. Median wages are less sensitive to outliers than mean wages and thus may be a better measure when income distributions are highly skewed, such as those in many developing countries. They are, however, only infrequently reported, thus limiting the sample severely. Mean wage data, by contrast, are relatively noisy and volatile, and inconsistently measured across countries (e.g., detrended in some countries but not in others). Also, mean wages typically correspond to average wages in manufacturing for males and females in full-time employment, even though in many countries, especially low-income countries, the manufacturing sector represents only a small part of the economy and women may represent only a small part of the workforce in these sectors. Nevertheless, to maximize data coverage, we calculate the ratio of minimum to mean wage as our baseline indicator.⁴

Unemployment Insurance

We construct two unemployment insurance (UI) indicators to capture different aspects of unemployment insurance systems:

• The level of UI benefits captures the generosity of the unemployment benefit system and is measured by the gross replacement rate (GRR), that is, the ratio of UI benefits a worker receives relative to the worker’s last gross earning.⁵ The database contains GRR measures for the first year of unemployment, the second year of unemployment, and the average of the two.

• The number of UI benefit recipients is calculated as the number of individuals who, at a given point in time, receive UI benefits. Relative to the number of unemployed, it can proxy the extent and reach, or exclusivity, of the UI system in a given country and thus provide complementary information to the generosity of the UI system.

To construct the GRR, we collect information on the earnings base, waiting period, rules of UI payment, maximum duration, and minimum and maximum payments (UI benefit ceilings). We also determine the year of introduction of the first legislation and the years of all consecutive reforms, and record the rules and procedures set out by each law, following six steps:

For information on the number of UI benefit recipients, we additionally rely on the following sources: national statistics offices and national statistical yearbooks, ministries of labor, social protection, and employment, social security administrations, labor funds, other bodies who administer the programs, ministries of finance and economy, national central banks, local research institutes, and national libraries wherever available.

Unemployment assistance is not generally considered part of UI, and we thus do not include such information in the calculations of GRR and coverage. However, some countries, namely, Australia, Hong Kong, and New Zealand, do not have an unemployment benefit system, but instead highly developed unemployment assistance schemes; in these cases, we calculate GRRs on the basis of unemployment assistance. In other countries, such as Chile (2002–05), and Venezuela (2002–05), reforms took place towards broader social insurance systems (see, e.g., Acevedo, Eskenazi, and Pagés, 2006). For example, Chilean unemployment insurance is based on two components: individual capitalization accounts, to which workers are contributing, and a common fund, to which the employer and the state are contributing. We compute the amount of benefits for reference individuals with the maximal legislatively set length of contributions to the individual savings accounts.

In most countries, we calculate GRRs on a de jure basis. However, this is possible only in countries where rules for UI payments are expressed in percent of previous earnings. In a number of countries, however, UI payments are set as a percentage of a minimum wage or a subsistence minimum, or as a flat rate payment. In these cases, we calculate GRRs on a de facto basis, as the ratio of these payments to previous earnings, proxied by the average wage in manufacturing. See Table 3 for the list of such cases. This procedure may affect comparability of the calculated GRRs across countries.

Table 3.

Countries in the Database with De Facto Gross Replacement Rates

Table 3.

Countries in the Database with De Facto Gross Replacement Rates

Albania (1993-2005)	Hong Kong (2001-2005)
Algeria (1994-2005)	Ireland (1980-2005)
Argentina (1985-1991)	Kyrgyzstan (1991-2005)
Australia (1980-2005)	Lithuania (1991-2005)
Chile (1985-2005)	New Zealand (1980-2005)
China (1999-2005)	Poland (1996-2007)
Estonia (1992-2002)	Portugal (1980-1984)
Finland (1985-2005)	Tunisia (1982-2005)
Georgia (1998-2005)	UK (1990-2005)

View Table

Table 3.

Countries in the Database with De Facto Gross Replacement Rates

Albania (1993-2005)	Hong Kong (2001-2005)
Algeria (1994-2005)	Ireland (1980-2005)
Argentina (1985-1991)	Kyrgyzstan (1991-2005)
Australia (1980-2005)	Lithuania (1991-2005)
Chile (1985-2005)	New Zealand (1980-2005)
China (1999-2005)	Poland (1996-2007)
Estonia (1992-2002)	Portugal (1980-1984)
Finland (1985-2005)	Tunisia (1982-2005)
Georgia (1998-2005)	UK (1990-2005)

A methodologically difficult area is that of UI coverage, conceptually, the fraction of unemployed individuals who collect UI benefits.⁶ While its calculation is straightforward, cross-country comparability is problematic. In most countries with UI systems in place, the number of UI benefit recipients (the numerator of the coverage index) is a highly accurate statistic: it is collected by the offices that effectuate the payments, based on officially claimed and received benefits, and is further aggregated by bodies administering the UI system, such as UI boards, national insurance institutes, or national employment offices. It is typically reported on a cumulative basis, such as the number of recipients for a given period of time, usually a month, a quarter, or a year.

By contrast, the number of unemployed individuals (the denominator of the coverage index), is usually measured with less precision, especially in non-OECD countries, and it is often particularly difficult to account for unofficial and hidden unemployment. Labor force surveys may also underestimate actual unemployment. For example, in many countries, labor force surveys focus on metropolitan regions where unemployment rates are often lower. Because the statistics of recipients cover the whole country, the UI coverage ratio can be implausibly high, exceeding one in some cases.

The number of unemployed is a statistic that is also conceptually different from the number of recipients: the former is usually given for a specific point in time, while the latter is reported on a cumulative basis, that is, as the number of all individuals who during a given year received UI benefits for any length of time.⁷ Lastly, countries also set different rules for UI payment, with, for example, some making UI payments even to partly-employed workers. These caveats imply that cross-country comparisons may not always be informative. However, to the extent that national definitions remain unchanged over time, the indicator can provide useful information on within-country dynamics. That said, we report both the number of recipients and the number of unemployed in the database to allow researchers to choose whichever variable best suits their purpose.

Employment Protection

The database contains two main indicators of employment protection legislation (EPL), reflecting advance notice requirements and legally mandated severance payments, for workers with 9 months, 4 years, and 20 years of experience, respectively. We report advance notice and severance pay requirements both in monthly salary equivalents and coded according to the OECD methodology. For their construction, we followed the same (six) steps as those for UI, based on the various EPL publications. In particular, the ILO Termination of Employment Legislation Digest, a database that describes EPL currently in place in a selection of countries, served as one of the main sources of information for EPL provisions. The main data sources for most of Africa, Asia, and the former Soviet Union were national labor codes and national legislation, some of which were obtained from the ILO Library Archives and the ILO NORMES Database, which are open to the public at the ILO head office in Geneva. For transition economies, we also relied on the Tonin (2007) database of EPL.

Unlike UI systems, EPL, in the form of either advance notice or severance payment requirements, has been in place in the vast majority of countries during 1980–2005. All of the EPL indices in the database are de jure, based on the provisions of legislation in place, such as labor codes, employment protection acts, and other types of laws⁸.

OECD Benefits and Wages, Minimum Wages, and Employment Protection Databases

These databases contain detailed information on all indicators that we are reporting for the period from 1960 to 2005, but restricted to OECD countries. The OECD databases are the main point of departure for our database. Specifically, for the OECD countries, we use the data directly from these datasets (except for the data on UI coverage, which we construct for both OECD and non-OECD countries). For all countries outside the OECD, we also apply the OECD data coding methodology in addition to providing the actual raw data. For example, some of the indicators, such as advance notice and severance payment, are part of the OECD EPL indices. Thus, an important contribution of our new dataset is its extention of three subject areas to non-OECD countries, especially lower-income countries, and the addition of information on UI coverage for all countries.

Social Security Programs throughout the World (SSPTW)

The SSPTW reports are descriptive in character, in contrast to the quantitative nature of our dataset. They contain information on UI systems (among other indicators) for most countries in our sample as well as additional ones. These reports describe mainly the current legislation, although they also provide the year of the first law for unemployment benefit provisions and the year of entry into force of current legislation. Our dataset uses the SSPTW as one of the main information sources for coding current UI regulations, and for determining whether regulations exist at all.

World Bank Doing Business (DB) Indicators

The DB database covers 181 countries, but provides information only starting in 2004. Among many other subjects, the DB database contains information on firing cost and on the difficulty of firing workers. The main difference from our new database is that the DB database is based on experts’ assessments of the severity of laws and regulations, and the coding of indicators is based in large part on survey questionnaires completed by local law firms. Partly reflecting their subjective nature, the DB indicators, especially those pertaining to the “Employing Workers” component, have been criticized (see Berg and Cazes, 2008, for a detailed discussion). By contrast, our database is a descriptive coding of the actual laws and regulations that are in place and does not take a stance on the desirability of a given level of regulation.⁹

Botero, Djankov, La Porta, Lopez-de-Silanes, Shleifer (2004)

This dataset covers 85 countries and a variety of indicators in the areas of employment laws; collective relations laws; and social security laws, partly overlapping with our database. However, it provides only a one-year snapshot (1997) of these regulations. By contrast, our database allows for the tracking of changes in labor market regulations over time.

Rama and Artecona (2002)

This database provides information for 121 countries, partly overlapping with our sample, during 1945-1999. It contains 44 labor market indicators, including the nominal minimum wage (in current US dollars); the initial UI benefit (in percent of earnings before job loss); the maximum duration of continuous unemployment benefits; and the mandatory severance pay after three years of employment (in months of salary). The data are reported in the form of five-year averages. While this aggregation was done deliberately, as many institutions are rigid and do not often change over time, it hinders the dating of reforms and cannot pick up rapid changes, such as those during 1990–2005 in many transition economies. Thus, Rama and Artecona’s (2002) and our database are complementary in that they cover similar regulations but using different approaches.

Other datasets

Several other datasets exist that are related to ours along various dimensions, including the LIS Comparative Welfare States Dataset (see Huber et al, 2004); the Fondazione Rodolfo Debenedetti Social Reforms Database; the Fraser Institute Economic Freedom of the World Annual Reports; and the Harvard Labor and Work Life Program’s Global Labor Survey (Chor and Freeman, 2004; Freeman, 2007). Interested readers are encouraged to review these databases for further detail.