Coursehero >> New York >> Cornell >> DUTCH 300

Course Hero has millions of student submitted documents similar to the one below including study guides, homework solutions, papers, exam answer keys and textbook solutions.

Strategy Employment Papers Imputation, estimation and prediction using the Key Indicators of the Labour Market (KILM) data set Gustavo Crespi Tarantino Employment Trends Unit Employment Strategy Department 2004/16 Employment Strategy Papers Imputation, estimation and prediction using the Key Indicators of the Labour Market (KILM) data set Gustavo Crespi Tarantino Science and Technology Policy Research (SPRU) University of Sussex United Kingdom Employment Trends Unit Employment Strategy Department 2004/16 Copyright page Preface This paper was prepared as background research for the World Employment Report 2004-05, Employment, Productivity and Poverty Reduction. The topic of this year s Report was chosen based on the observation that it is not simply the lack of employment that leads to poverty, but rather the lack of decent and productive employment. In many parts of the developing world the poor are in fact employed, but employed in such poorly paid conditions that they and their families live on less than US$1 a day per person. Thus, unemployment is only the tip of the iceberg of the decent work deficit. The Report concludes that not only do we need more jobs, but more productive jobs jobs that allow workers to lifts themselves and their families out of the vicious cycle of poverty. The background papers commissioned for this Report provide an overview of the important aspects involved in the links between employment, productivity and poverty reduction in both developing and developed economies. The papers were commissioned from experts in the field as well as various departments within the ILO and discuss different avenues through which poverty can be reduced, as well as the trade-offs that must be made in order to strike the right balance between productivity, employment and income growth. The research involves macroeconomic, sectoral and case study analysis that has helped form the basis of the chapters in the Report. Based on the research from these background papers the Report concludes that increasing the opportunity for decent and productive work is an important channel towards achieving a fairer globalization, and is vital for poverty reduction. Duncan Campbell Director a.i. Employment Strategy Department Contents 1. 2. 3. Introduction................................................................................................................................................ 1 Labour Market Indicators ........................................................................................................................ 1 Methodology ............................................................................................................................................... 2 3.1 Preliminary Considerations.............................................................................................................. 3.2 Ad Hoc Country-Level Interpolations.............................................................................................. 3.3 Analysing the Missingness Mechanism......................................................................................... 3.4 Estimation and Prediction of Unemployment Figures ..................................................................... Results......................................................................................................................................................... 2 3 4 6 7 4. 5. 6. 4.1 Country-Level Interpolations ............................................................................................................ 7 4.2 Determinants of the Response Rates ................................................................................................. 8 4.3 Imputation and Prediction of Unemployment Figures ...................................................................... 9 4.4 Model Evaluation............................................................................................................................ 11 Employment Distribution by Economic Activity ................................................................................... 13 Conclusions................................................................................................................................................ 18 Bibliography ........................................................................................................................................................ 20 Appendix Tables Table 4.1: Table 4.2: Table 4.3: Table 4.4: Table 4.5: Table 4.6: Table 4.7: Table 5.1: Table 5.2: Table 5.3: Table 5.4: Table 5.5: Table A.1: Table A.2: Table A.3: Table A.4: Table A.5: Table A.6: Table A.7: Table A.8: Table A.9: Table A.10: Table A.11: Table A.12: Table A.13: Table A.14: Table A.15: Table A.16: Table A.17: Table A.18: Table A.19: Table A.20: Table A.21: Figures ........................................................................................................................................................ 21 Response rates by subregion ............................................................................................................ 7 Response rates by subregion, after interpolations ............................................................................ 8 Determinants of the response probability......................................................................................... 9 Total unemployment figures at world level .................................................................................... 10 Total unemployment rates at world level ........................................................................................ 11 Imputed unemployment rates according to different methodologies: Population average.............. 12 Imputed unemployment rates according to different methodologies: Population standard deviation.......................................................................................................................................... 13 Response rates by subregion before interpolation........................................................................... 15 Response rates by subregion after interpolation.............................................................................. 16 Determinants of response probability: Employment distribution by economic sector.................... 16 Employment distribution by economic sector ( 000)...................................................................... 18 Employment distribution by economic sector (%).......................................................................... 18 Total Male and Female Unemployment ( 000) ............................................................................... 21 Total Youth Male Unemployment ( 000) ....................................................................................... 22 Total Adult Male Unemployment ( 000) ........................................................................................ 23 Total Youth Female Unemployment ( 000).................................................................................... 24 Total Adult Female Unemployment ( 000)..................................................................................... 25 Male - Female Unemployment Rates (%)....................................................................................... 26 Youth Male Unemployment Rates (%) .......................................................................................... 27 Adult Male Unemployment Rates (%)............................................................................................ 28 Youth Female Unemployment Rates (%) ....................................................................................... 29 Adult Female Unemployment Rates (%) ....................................................................................... 30 Male - Female Employment-to-Population (%) ............................................................................. 31 Youth Male Employment-to-Population (%) ................................................................................. 32 Adult Male Employment-to-Population (%).................................................................................. 33 Youth Female Employment-to-Population (%) ............................................................................. 34 Adult Female Employment-to-Population (%) ............................................................................... 35 Total Employment by Economic Sectors: Agriculture ( 000)........................................................ 36 Total Employment by Economic Sectors: Industry ( 000).............................................................. 37 Total Employment by Economic Sectors: Services ( 000) ............................................................. 38 Total Employment by Economic Sectors: Agriculture (%)............................................................. 39 Total Employment by Economic Sectors: Industry (%).................................................................. 40 Total Employment by Economic Sectors: Services (%) ................................................................. 41 Figure 5.1 Employment shares and per capita income by economic sector 14 1. Introduction Almost all empirical analysis has the problem of missing data, especially survey analysis, market studies and social science research.1 This issue is so persistent, and its consequences so serious, that a large and growing literature has emerged on how to deal with the problem.2 The Key Indicators of the Labour Market (KILM) data set3 produced by the ILO is a timely and significant advance in coordinating data and improving comparability on a large set of indicators on labour market conditions in ILO member countries. This information is highly relevant for policy-makers and researchers alike, on issues that include employment, education, productivity, economic growth, poverty and gender discrimination. It is essential to generate such information at the micro (country) level, not only to produce consistent aggregated statistics, but also because an in-depth understanding of labour market issues necessitates the study of the individual structural characteristics of each national economy. To achieve this goal, the Employment Trends Team of the ILO s Employment Strategy Department has made an intensive effort to collect information on labour market conditions from all sources in ILO member countries, to analyse these data and to produce comparable cross-country statistics. Given its complexity and ever-increasing geographical range, it is not surprising that the KILM data set is affected by the problem of missing data. Large gaps exist in the information submitted, particularly from developing countries. The ideal solution is that all countries collect and submit the same data, but this is a learning and investment process that will take time to happen. Meanwhile, the question is whether it is possible, in the short-term, to use the available information in order to monitor trends in labour market conditions at least at global and regional levels. The aim of the present report is to apply methods that make use of the data already collected by the ILO to estimate current labour market indicators for those countries where information is not yet available and to predict future trends for world labour markets. This report describes the main methodological approaches used in the research, summarizes the procedures applied to the data set and presents the results of the projections corresponding to unemployment, the employment-to-population ratios and the distribution of employment by economic sectors. 2. Labour market indicators The methods described and applied in this paper aim to generate statistics for the following labour market indicators: 1) 2) Total unemployment (both rate and count). Total unemployment stratified by sex and, within sex, stratified by age (both rates and counts). Unemployment by age will be decomposed using two broad categories: individuals of 15-24 years old or the closest available and individuals of +25-64 years old. Data are presented on both rates and counts. Total employment (computed as the residual between labour force and total unemployment). 3) 1 Missing data are questions without answers, variables without observations, or units that refuse to respond. For a summary see LITTLE, R., and D. RUBIN (1987) and SCHAFER, J. (1997). 3 See ILO (2003): Key Indicators of the Labour Market, 3rd edition, for more details. 2 2 4) The allocation of total employment across different economic sectors. (using 1 digit ISIC, which corresponds to the following activities: agriculture, manufacturing and services both shares and counts). The sample framework for the estimations is given by the labour force projections (total, by sex and age) generated by the ILO Bureau of Labour Statistics (the LABPROJ files).4 Hence, it is not part of the current methodology to predict the labour force; this is taken as given by the team responsible for KILM. The time frames for the predictions are the periods 2002-2003-2004 and 2015. 3. Methodology Preliminary considerations In order to predict the future conditions of world labour markets, some statistical correlation between the different labour market indicators and the macroeconomic context must be identified. Broadly speaking, three different information sets need to be assembled: labour force figures, macroeconomic conditions and unemployment data. In this exercise, the information about labour force figures was taken as given in the LABPROJ data set, while the historical and projected values for economic growth were taken from the International Monetary Fund5 and the World Bank.6 The focus here is on the statistical procedures applied to the incomplete unemployment figures in the KILM data set. Some preliminary work identified three different issues with the reported unemployment figures. First, not all countries submitted the information with the detail required for analysis.7 Second, a large group of countries do not report unemployment figures.8 Third, even if a complete data set were available, the issue of the heterogeneity among reporting countries persists not only in terms of the idiosyncrasies of their labour market conditions, but also in terms of how they collect and process the raw information.9 If the results of the statistical analysis are to be fully representative of the world labour market situation, these points must be accounted for. The methodology applied in this research is developed in three steps. First, the issue of incomplete information for the unemployment sub-components is treated by using ad hoc country-specific imputation techniques. This makes it possible to preserve intact the richness provided by the heterogeneity of the data set and also to maintain consistency where countrylevel statistics exist. Second, the missingness data mechanism of the full data set is analysed. This approach not only tests whether non-reporting countries are statistically different from reporting ones (these latter cannot be considered a simple random sample of See SCHAIBLE, W. and R. MAHADEVAN-VIJAYA (2002): "World and Regional Estimates for Selected Key Indicators of the Labour Market," Employment Paper No. 2002/36, for a description of the labour force projections conducted at the ILO. 5 http://www.imf.org/external/pubs/ft/weo/2003/02/data/index.htm. 6 World Development Indicators, various years. 7 Information was required not only for total (male and female) unemployment rates but also for each micro component according to sex and age (two different age categories were used: youth workers under 25 years and adult workers over 25 years). 8 For total unemployment rates the response rate is slightly higher than 50 per cent but for some of its subcomponents the rate is only about 30 per cent. 9 Some countries report information from household surveys or population censuses, while other countries report information generated by official employment agencies. Some countries refer to the total labour force, others only to the civil labour force. 4 3.1 3 the total population), it also builds different weights in order to balance the sample of reporting countries and to generate a working sample that looks closer to a random selection of the whole population. The third and final step in the methodology is to apply panel data techniques to the observed sample in order to estimate and predict future labour market conditions. Panel data estimation techniques control for the massive heterogeneity underlying the data. In order to keep consistency between total unemployment and its different subcomponents, it was decided to follow a bottom up strategy. This means that the primary unit of analysis is the lowest possible disaggregated sub-components of unemployment: Youth Male Unemployment, Adult Male Unemployment, Youth Female Unemployment and Adult Female Unemployment. This strategy is very demanding in terms of data availability for the different countries, but it produces more consistent information between strata. Ad hoc country-level imputations As outlined above, the first step in the methodology was to address the issue of lack of complete information for the unemployment sub-components within the set of reporting countries. Although many countries report total unemployment rate figures for several of the years, unemployment rates for the different sub-components are not reported for all these years. There is some implicit statistical correlation between total unemployment rate and its components at country level that can be exploited in order to recover the basic information. Imputation of missing unemployment sub-components followed two procedures: 1) A panel data set of roughly ten years of information for each country was first assembled. Where information on unemployment sub-components was missing for some years, information from complete years was used to fill the gaps. More specifically, for the observed year(s) different sub-components to total ratios were computed and the median of these ratios was used to impute for the years with missing sub-components. 2) When information on sub-components was missing for all the years in one reporting country (but not the information for total unemployment) the gaps were imputed using a similar procedure, but with the ratios now computed at regional and subregional levels. It is important to note that simple linear interpolations do not work very well in this case. The dependent variable under analysis, the unemployment rate, is censored at the interval [0, 1]; as a consequence simple linear interpolations can potentially generate out of range imputed values. In order to control for this, the different unemployment rates were transformed using a logistic function and the adjustment factors described in 1) and 2) above were defined in terms of differences. These adjustment factors were then added to the (logistically transformed) total unemployment rates in order to obtain (transformed) imputed values for the missing unemployment sub-components. Finally, the inverse transformation was applied and the original unemployment rates were recovered. More formally, we can define a transformed dependent variable as follows: y T Yitk = ln itk 1 y itk 3.2 (1) where yitk is the observed unemployment rate for sub-component k in country i and period t. There are four sub-components (youth male unemployment, adult male unemployment, youth 4 female unemployment and adult female unemployment). Let us also define a transformed independent variable such as: y YitT = ln it 1 y it (2) Where yit is the observed total unemployment rate in country i and period t, the adjustment factor is then defined as: T AFi = Med Yitk YitT ( ) (3) It is then possible to recover the missing unemployment rate for the k sub-component with the condition that the total unemployment rate is observed as follows: ~T Yitk = AFi + YitT T Yitk = missing (4) When a country reported only total unemployment rates, but not sub-components, we used subregional level adjustment factors. Analysing the missingness mechanism A large proportion of the countries included in the KILM data set do not report unemployment figures, which raises some concern about the lack of representativeness of the sample. Are the reporting countries sufficiently similar to be used to impute the incomplete unemployment figures? Following Horowitz and Manski (1998), each country in the KILM data set can be characterized by a vector (yit, xit, wit, rit), where y is an outcome of interest (the unemployment rate), x is a set of covariates that determines the value of the outcome and w is a set of covariates that affects the probability of the outcome being observed. Finally, r is a binary variable indicating a missing response as follows: 1 if i reports rit = 0 if i is missing 3.3 (5) The focus of the problem is estimating conditional expectations for unemployment rates of the form E[g ( yit ) | xit A] where g(.) is a specified real-valued function of outcome yit and A is a specified set of values of the covariates xit. Following from (5), rit=1 indicates that the set (yit,xit) is fully observed and rit=0 that data on yit are missing. The vector of covariates wit, which is always observed, is used to balance the observed sample of countries by computing weights. These covariates include a set of country-specific characteristics such as economic growth, per capita GDP, population and membership in the Heavily Indebted Poor Countries Initiative.10 More specifically, by conditioning on the given set of covariates wit and using logistic regression, it is possible to estimate each country s probability of reporting unemployment 10 The principal objective of this UN programme is to bring the debt burden of the HIPC countries to sustainable levels, subject to satisfactory policy performance, so as to ensure that adjustment and reform efforts are not put at risk by continued high debt and debt service burdens. One by-product of the initiative is that national statistics offices in HIPC countries are required to collect fuller information and to strengthen their data capabilities. 5 figures. Let us assume that there is a linear function connecting some unobserved index value of reporting unemployment figures with the set of covariates: ' rit* = wit + it (6) where each country reports if this index value is positive ( rit* > 0 ). From (6) and using (5) it is possible to model the probability of reporting unemployment figures as: ' Pi = P(rit = 1) = P it > wit ' = 1 F wit ( ( ) ) (7) where F is the cumulative distribution function of it. If this distribution function is symmetric, (7) can be rewritten as: ' Pi = F wit (8) () Since the observed rit are just realizations of a binomial process with probabilities given by (8), the likelihood function of this problem is given by: L = Pi (1 Pi ) ri =1 ri =0 (9) The functional form for F in (9) will depend on the assumption made about the error term it. In this application we assume that the cumulative distribution of this term is a logistic (a logit model is thus estimated). That is: ' F w it = ( ) ' exp w it ' 1 + exp w it ( ( ) ) (10) After estimating (10) we can compute the predicted response probabilities for each individual country in the data set. These predicted probabilities are then used to compute weights defined as: sit (w) = P(rit = 1) P(rit = 1 | wit , ) (11) The key point here is that according to (11), the weights are computed as the ratio between the proportion of non-missing observations in the sample and the reporting probability attached to each country in each year. In this way, the influence in the sample of those reporting countries that are more similar (according the covariates wit) to the missing ones is inflated, while the importance of those that are quite different is diminished. As a result the weighted sample looks more similar to the theoretical population framework than the unweighted sample of reporting countries. After computing the weights, the results of interest E[g ( yit ) | xit A] are estimated by the weighted average s(wit )g ( yit ) where N1 is i N1 the set of reporting countries. 6 3.4 Estimation and prediction of unemployment figures This section deals with the specification of the function E[g ( yit ) | xit A] . Here the critical issue is the treatment of the unobserved heterogeneity of the various countries. Panel data techniques, which take the unobserved heterogeneity into consideration, are used. In order to apply these techniques, a data set containing information for each country over a tenyear period (with gaps filled as described in section 3.2) was built. All regressions were estimated using fixed-effect methods with the sample of the reporting countries weighted to consider the non-response bias. That is, we estimated the following linear model: y T Yitk = ln itk 1 y itk ' = i + xit + it (12) where yitk is the observed unemployment rate for sub-component k in country i and period t and xit is a set of covariates explaining the unemployment rate. In this research this set of covariates is based only on GDP growth rates and some time dummies (but only for regions where there is clear evidence of structural change). The constant i is country-specific, capturing all the persistent idiosyncratic factors governing the unemployment rate in each country. This model can be estimated using the sample of respondent countries by introducing a set of country-level dummy variables as follows: Y T = X + d1 1 + d 2 2 + L + d N1 N1 + (13) Model (13) is estimated by weighted least-square methods, the weights being those computed in section 3.2. These weights are first normalized using the expression: s(wit ) = * s(wit ) N1 s(wit ) (14) After estimation the model can also be used for imputation and prediction. Although it is not strictly necessary to carry out imputations for the non-respondent countries, doing so can be useful to produce a complete data set that can be used to compute additional statistics about labour market conditions or to generate different regional and subregional aggregations. However, it is not straightforward to use a fixed-effect model for imputing missing countries and whatever fixed effect is predicted for the non-respondent countries would be arbitrary. An intuitive approach in the context of this study is to impute a fixed effect for the non-respondent countries based on the weighted average of the fixed effects for respondent countries. That is, given N1 mutually exclusive and exhaustive country dummies, we identify a weighted average fixed effect by choosing the intercept that makes the prediction calculated at the (weighted) means of the independent variables equal to the (weighted) mean of the dependent variable: ___ Y = X + ___ ___ (15) This newly complete data set was used in order to create the subregional and regional aggregates.11 All the models were estimated at subregional level. 11 The standard errors were also computed. However, these are also underestimated because they do not take account of the uncertainty associated with the estimation of the fixed effects. 7 4. Results Three different sets of results are obtained from the methodological approach followed in this paper: the results related to country-level interpolations for the respondent countries; the results of identification of the missingness mechanism; and, finally, the statistics from the panel data model. We present each in turn. Country-level interpolations Table 4.1 summarizes the response rates by subregion. The subregions are as defined in the KILM data set and a complete response means that the country has reported both total unemployment rates and the four corresponding sub-components. Table 4.1 shows that the global response rates are quite low, averaging 30 per cent for the entire period. The global response rate is higher in the middle of the time frame, but deteriorates very quickly towards the end of the period suggesting serious lags in how member countries collect, process and report their raw information. The response rate varies even more across subregions. Response rates are near 100 per cent for the developed world but are much lower for the other subregions. South America is the second highest subregion, with an average response rate of 60 per cent. Relatively high response rates are also found in Eastern Europe and Asian countries, in contrast to the low response rates observed in the poorest subregions of Africa. 4.1 Table 4.1: Response rates by subregion KILM Subregion Major Europe Major Non-Europe Other Europe Eastern Europe Baltic States CIS Melanesia Eastern Asia South-Central Asia South-Eastern Asia Central America South America Eastern Africa Middle Africa Southern Africa Western Africa Middle East North Africa Total Source: ILO, KILM, 3rd edition. 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Total 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.43 0.13 0.09 0.47 0.58 0.00 0.00 0.00 0.00 0.07 0.00 1.00 1.00 0.00 0.17 0.00 0.08 0.00 0.43 0.25 0.18 0.37 0.67 0.00 0.00 0.00 0.00 0.07 0.00 1.00 1.00 0.00 0.42 0.00 0.00 0.00 0.43 0.25 0.18 0.47 0.67 0.00 0.00 0.00 0.00 0.07 0.00 1.00 1.00 0.00 0.42 0.00 0.00 0.00 0.43 0.25 0.27 0.42 0.67 0.00 0.00 0.20 0.00 0.07 0.00 1.00 1.00 0.00 0.50 0.00 0.17 0.00 0.43 0.25 0.18 0.68 0.75 0.00 0.00 0.00 0.00 0.13 0.00 1.00 1.00 0.00 0.50 0.33 0.08 0.00 0.43 0.13 0.36 0.63 0.75 0.00 0.00 0.00 0.00 0.13 0.00 1.00 1.00 0.00 0.50 1.00 0.00 0.00 0.43 0.13 0.27 0.53 0.58 0.00 0.00 0.40 0.00 0.07 0.00 1.00 1.00 0.00 0.58 1.00 0.08 0.00 0.43 0.25 0.27 0.63 0.75 0.00 0.00 0.20 0.00 0.07 0.40 1.00 1.00 0.00 0.42 0.67 0.25 0.00 0.14 0.13 0.18 0.53 0.67 0.06 0.00 0.20 0.00 0.07 0.40 1.00 1.00 0.50 0.50 1.00 0.17 0.00 0.43 0.38 0.27 0.37 0.58 0.00 0.00 0.00 0.00 0.07 0.00 1.00 1.00 0.50 0.67 0.67 0.08 0.00 0.43 0.00 0.09 0.21 0.50 0.00 0.00 0.00 0.00 0.07 0.00 0.95 1.00 0.00 0.42 0.00 0.00 0.00 0.14 0.00 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 1.00 0.08 0.42 0.39 0.08 0.00 0.38 0.18 0.20 0.45 0.60 0.01 0.00 0.08 0.00 0.07 0.07 0.26 0.28 0.30 0.31 0.35 0.35 0.34 0.38 0.34 0.34 0.29 0.17 0.31 The definition of response used in Table 4.1 is rather precise in that it requires each country to report information for all the unemployment sub-components. Some countries report information for total unemployment for all the years, but report on sub-components for 8 only some years. As explained in Section 3 on methodology, we use country-level ratios in order to recover the sub-components for the missing years in a given country. This procedure, which makes more efficient use of the reported information, increases the number of reporting countries. The results of these adjustments on the response rates are summarized in Table 4.2. Table 4.2: Response rates by subregion, after interpolations KILM Subregion Major Europe Major Non-Europe Other Europe Eastern Europe Baltic States CIS Melanesia Eastern Asia South-Central Asia South-Eastern Asia Central America South America Eastern Africa Middle Africa Southern Africa Western Africa Middle East North Africa Total Source: Author s calculations. 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Total 1.00 1.00 1.00 0.75 0.67 0.33 0.33 0.71 0.25 0.27 0.79 0.75 0.13 0.00 0.20 0.00 0.20 0.40 1.00 1.00 1.00 0.75 1.00 0.67 0.33 0.71 0.25 0.36 0.74 1.00 0.25 0.00 0.20 0.00 0.13 0.40 1.00 1.00 1.00 0.83 1.00 0.75 0.33 0.86 0.25 0.36 0.79 0.92 0.19 0.13 0.20 0.00 0.13 0.40 1.00 1.00 1.00 0.83 1.00 0.75 0.33 0.86 0.25 0.36 0.79 0.92 0.25 0.00 0.20 0.00 0.13 0.40 1.00 1.00 1.00 0.92 1.00 0.75 0.33 0.86 0.25 0.36 0.79 0.92 0.19 0.00 0.40 0.00 0.33 0.40 1.00 1.00 1.00 0.92 1.00 0.75 0.00 0.86 0.38 0.45 0.79 0.92 0.25 0.00 0.40 0.06 0.33 0.40 1.00 1.00 1.00 0.92 1.00 0.75 0.00 0.86 0.25 0.45 0.79 0.92 0.19 0.00 0.60 0.06 0.40 0.60 1.00 1.00 1.00 0.92 1.00 0.67 0.00 0.86 0.25 0.45 0.84 0.92 0.13 0.00 0.60 0.00 0.33 0.60 1.00 1.00 1.00 0.92 1.00 0.67 0.00 0.86 0.38 0.45 0.84 0.92 0.19 0.00 0.60 0.00 0.47 0.60 1.00 1.00 1.00 0.92 1.00 0.67 0.00 0.86 0.38 0.55 0.68 0.75 0.06 0.00 0.60 0.00 0.53 0.60 1.00 1.00 1.00 0.92 1.00 0.67 0.00 0.86 0.13 0.55 0.63 0.75 0.00 0.00 0.20 0.00 0.47 0.20 1.00 1.00 1.00 0.92 1.00 0.67 0.00 0.86 0.13 0.36 0.53 0.58 0.00 0.00 0.00 0.00 0.27 0.20 1.00 1.00 1.00 0.88 0.97 0.67 0.14 0.83 0.26 0.42 0.75 0.85 0.15 0.01 0.35 0.01 0.31 0.43 0.47 0.52 0.54 0.54 0.56 0.58 0.58 0.57 0.59 0.56 0.51 0.46 0.54 On this basis, global response rates are much higher reaching a value of 54 per cent for the whole period. The response rates for each subregion also increase. However, response rates are still very low for some of the sub-Saharan Africa subregions. Therefore, in the following sections all these subregions are merged and no attempt is made to report results for individual subregions. Similarly, the Melanesia subregion is merged with the SouthEastern Asia subregion. Determinants of the response rates Table 4.3 shows the results of the logit regressions on the determinants of reporting information. The different countries have been allocated to the different subregional groups established previously. It is important to note that there are no results for the developed countries as the same ones reported full information. Four explanatory variables were used in each case: the (log) per capita income; the gross domestic product growth rate; the size of the country measured by total population; and membership in the HIPC programme.10 The last two rows Table 4.3 summarize goodness-of-fit of the different models. In general, when according the pseudo-R2, the models predict relatively well. More important are the likelihood ratio tests on global significance of the model. If these tests suggest that the explanatory variables are (jointly) non-significant in explaining the response rates, it would 4.2 9 be possible to infer that countries are missing randomly (at least within particular subregions). In this case, it would be valid to consider the reporting countries as being a random sample of the total. The values of the tests suggest that this hypothesis can be strongly rejected for all subregions except Central Asia where it was not rejected (but only marginally). As a consequence it is possible to infer that the four explanatory variables used in the analysis are correlated with the probability of response. Indeed, in five of the nine regions a growth in the per capita income of countries increases the response probabilities: in three subregions the GDP growth rates were negatively correlated with the response rates; in five subregions the more populous countries were among those with the highest response rates and, finally, HIPC programme membership was significant for only three subregions (positively associated with the response probabilities in CIS countries and negatively correlated for Africa and South America). Table 4.3: Determinants of the response probability Eastern Europe CIS Eastern Asia Central Asia South Central Asia Central America South America SubSaharan Africa North Africa Middle East PCGDP Growth Size HIPC Constant N Pseudo-R2 LR Chi2 0.267 (0.220) -0.328 (3.42)** 0.718 (1.700) 0.749 (0.920) 0.013 (0.360) 3.586 (4.75)** 2.485 (2.61)** 29.367 (.) 3.239 (0.000) 3.577 (0.010) 80.155 (0.010) -312.24 (.) 72 0.99 70.9** 2.488 (2.74)** -0.581 (2.33)* 0.238 (1.750) 1.402 (3.87)** -0.142 (2.49)* 0.573 (5.61)** 0.496 (0.750) 0.000 (3.71)** 0.085 (1.360) 0.001 (3.54)** 0.000 (0.050) 0.205 (0.650) 0.002 (2.28)* -7.650 (2.53)* 0.001 (6.80)** -0.011 (0.580) 0.000 (5.35)** -0.912 (2.68)** -2.668 (4.89)** 528 0.28 137.5** 0.000 (4.07)** -0.030 (1.240) 0.000 (5.33)** -1.543 (1.410) -2.653 (3.96)** 240 0.22 74.5** -8.958 (0.710) 156 0.34 37.1** -36.835 (3.94)** 144 0.37 39.1** -20.304 (2.74)** 84 0.21 18.51 -15.961 (4.69)** 168 0.26 58.8** -2.067 (2.02)* 204 0.18 34.4** 3.006 (0.600) 132 0.83 67.1** Notes: Absolute value of the Z statistic is in parenthesis. (*) significant at 5%, (**) significant at 1%. N = No. of observations. In sum, the largest and wealthiest countries show the highest response rates, although not necessarily the highest growth rates. This means that the responding countries cannot be considered to be a random sample of the total subregions. A simple imputation that ignored this issue would bias the results. We therefore decided to use the results of the logit models to predict the response probability for each country and to use this probability to create weights in order to re-balance the sample. 4.3 Imputation and prediction of the unemployment figures This section focuses on the results of the imputation and prediction models. All the models are weighted fixed-effect regressions estimated for each subregion as defined in Section 3 on methodology. Complete results are presented in Appendix tables A.1-A.15. The world-level trends are summarized below. Long-run trends were estimated on the assumption 10 that each country would continue to grow during the forecast period at its 1990s median growth rate. Table 4.4 presents total unemployment figures for the total labour force and its four different sub-components. The model produces a total of 185 million unemployed in 2003; this number is predicted to increase to 206 million by 2015. An important share in this grand total is youth unemployment, which in 2003 accounted for about 40 per cent of total unemployment. Table 4.4: Total unemployment figures at world level Year 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 MFU 153402 158310 159479 168592 176304 171825 173624 181558 184828 185304 206097 Sd 1388 1332 984 939 1033 956 975 1012 1653 1655 1693 YMU 44265 45103 44430 46889 49382 48380 48917 50948 51805 52315 56323 Sd 816 758 503 485 559 476 498 499 617 621 597 AMU 45500 47309 47815 50647 53032 51229 52188 54670 55666 55633 64974 Sd 841 835 682 645 678 674 673 706 1146 1147 1227 YFU 29902 30519 30788 32371 33519 33428 33734 34927 35583 35522 37379 Sd 560 509 334 322 369 310 332 336 474 473 421 AFU 33735 35379 36446 38685 40371 38789 38785 41013 41775 41834 47420 Sd 492 494 372 356 399 371 374 404 902 902 909 Note: MFU (male/female unemployment), Sd (standard deviation); YMU (youth male unemployment); AMU (adult male unemployment); YFU (youth female unemployment); AFU (adult female unemployment). Table 4.5 summarizes the world-level results of unemployment rates. During the second half of the 1990s global unemployment rates show an upward trend. Indeed, the world unemployment rate of 5.9 per cent in 1995 reached a peak of 6.3 per cent in 1999. This is a consequence of the accumulated impact of the Asian and Russian crises. There is a decline in this indicator during the next two years with unemployment rates dropping to 6.0 per cent in 2001. From 2001 onwards, and as a consequence of the slowdown in growth rates in developed economies, unemployment rates again rose, to 6.2 per cent in 2003. The long-run prediction is that if all the countries manage to grow at their 1990s medium growth rate, the global unemployment rate will decline to 5.9 per cent. The various sub-components of unemployment figures yield two interesting results. First, while adult (both male and female) unemployment rates move according to global trends, the youth strata show ever-increasing unemployment rates. This finding suggests that growth rates are not sufficient to absorb new cohorts of workers into the labour markets, leading to a worsening of the employment situation for these population groups. The prospects are worse for the female than for the male youth strata. By comparing across the different sub-components it is possible to demonstrate the persistence in the gaps in the unemployment rates between youth and adult cohorts. The ratio of youth unemployment rates to adult unemployment rates is, on average, about 3 to 1 and slightly growing over time. 11 Table 4.5: Total unemployment rates at world level Year 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 MFU 5.9 6.0 5.9 6.2 6.3 6.1 6.0 6.2 6.2 6.1 5.9 Sd 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 0.0 YMU 13.0 13.3 13.1 13.8 14.3 14.0 14.0 14.4 14.4 14.3 14.4 Sd 0.2 0.2 0.1 0.1 0.2 0.1 0.1 0.1 0.2 0.2 0.2 AMU 3.7 3.8 3.8 3.9 4.0 3.8 3.8 4.0 4.0 3.9 3.8 Sd 0.1 0.1 0.1 0.1 0.1 0.1 0.0 0.1 0.1 0.1 0.1 YFU 12.1 12.4 12.6 13.3 13.8 13.6 13.7 14.1 14.1 13.9 14.0 Sd 0.2 0.2 0.1 0.1 0.2 0.1 0.1 0.1 0.2 0.2 0.2 AFU 4.2 4.3 4.3 4.5 4.6 4.3 4.2 4.4 4.4 4.3 4.2 Sd 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 0.1 Note: MFU (male/female unemployment), Sd (standard deviation); YMU (youth male unemployment); AMU (adult male unemployment); YFU (youth female unemployment); AFU (adult female unemployment). It is important to note that while all tables include the corresponding standard deviations for the predictions, these statistics are severely underestimated. This is because the only uncertainty included in the predictions is that related to the unknown slope parameter in the regressions (that is, the partial correlation between unemployment rates and economic growth). There are, obviously, other important sources of uncertainty related to the estimation of country-level fixed effects and the possible errors involved in within-country interpolations. These standard errors should be considered inferior limits of uncertainty. One question remaining is how to evaluate the predictions generated by this methodology. This issue is not trivial, because whatever method is used to generate imputations and predictions should be compared with some complete data but this benchmark for the total world clearly does not exist. However, we can advance further if we restrict our analysis to a subset of cross-sections with relatively complete information: the developed countries. 4.4 Model evaluation This section summarizes the results of different exercises (carried out by simulation) aimed to evaluate the methodology applied during this research. The analysis starts by defining the set of developed countries as our complete information data set. Using this information several missing values are first generated. Different methodologies based on the remaining observed part of the data set are then used to impute the missing portions. The imputed data set is compared with the complete information data set in order to evaluate the different methodologies. Evaluation requires a definition of both the missing-value generating process and the approaches to be used during the imputations. With regard to the former, one significant result from section 4.3 is that, in many regions of the world, one important predictor of the missing-value patterns has been the per capita income of each country. Indeed, the poorest countries have a lower response probability. Using this result it is possible to specify the following hypothetical missing-values mechanism for developed countries: 12 0.95 0.60 P( yit = missing | wit ) = 0.35 0.20 if wit Q25 (wit ) if Q25 (wit ) < wit Q50 (wit ) if Q50 (wit ) < wit Q75 (wit ) if Q75 (wit ) (16) where yit means the total unemployment rate of country i at time t and wit is the corresponding per capita income. Equation (16) indicates that in any given year, countries in the 1st quartile of per capita income distribution will have a response probability of 5 per cent, countries in the 2nd quartile will show a response probability of 40 per cent, and those in the 3rd quartile a 65 per cent response probability. The response probability of the wealthiest countries will be 80 per cent. This missing-value generating process reproduces the previous result that the response probability is positively correlated with the per capita income. In addition, the probabilities were allocated in such a way as to produce a global response rate of about 50 per cent similar to the response rate in the real data. The now incomplete data set produced with equation (16) will be used in order to evaluate four different imputation approaches. The first two approaches are the two common methodologies already applied to the KILM data set.12 The first method is imputation using mean replacement. Here, the missing values in each year are replaced by the mean of the reported data in that year. The second approach is based on regression. Here, an unweighted least-square regression of the (transformed) unemployment rates on growth rates is used in order to impute the missing values. In addition to these two approaches, two other methods were tested. The third model evaluated also uses a least-square regression, but this time the observed part of the data set is weighted in order to take into account the response probabilities of the different countries. The weights are computed as in section 3.3. The fourth and final method to be evaluated is the one applied in this paper where not only are different weights used, but also fixed effects are introduced in the observed part of the data set. The results of these different methodologies can be seen in Table 4.6. Table 4.6: Imputed unemployment rates according to different methodologies: Population average Method Mean Replacement Unweighted Regression Weighted Regression Weighted Fixed Effects Complete Data Average 6.25 5.89 6.27 6.91 7.18 Bias 13.0% 18.0% 12.7% 3.8% Inferior Limit 5.94 5.61 5.94 6.41 Superior Limit 6.54 6.16 6.59 7.33 The results in Table 4.6 compare the averages over 1,000 simulations for the different imputation methodologies with the true (average) unemployment rates for developing countries. This true figure was 7.18 per cent over the entire period 1990-2002. All the methodologies tend to underestimate this number. Using mean replacement the imputed See SCHAIBLE, W. (2000): "Methods for Producing World and Regional Estimates for Selected Key Indicators of the Labour Market," Employment Paper for a description. 12 13 average is 6.26 per cent (a bias of 13 per cent), while pooled regression produces even worse results of average unemployment at 5.89 per cent (an 18 per cent underestimation). However, weighting significantly improves the performance of the pooled regression, leading to an average of 6.27 per cent (a bias of 12.7 per cent). However, the combination of weights plus fixed effects clearly produces the best results. Indeed, this approach produces an average of 6.91 per cent a bias of only 3.8 per cent. Inspection of the 10 per cent significance confidence intervals also confirms the advantage of using weights in combination with fixed effects: this is the only methodology where the true population parameter can be included within the confidence intervals, indicating that the imputed average is not statistically different from the true population average or that the bias is not statistically significant. Table 4.7: Imputed unemployment rates according to different methodologies: Population standard deviation Method Mean Replacement Unweighted Regression Weighted Regression Weighted Fixed Effects Complete Data Average 2.36 2.27 2.25 3.30 3.87 Bias 0.39 0.41 0.42 0.15 Inferior limit 2.11 2.03 2.02 2.69 Superior limit 2.62 2.52 2.50 4.17 Table 4.7 summarizes the results regarding the spreads of different imputed distributions. As expected, all the approaches tend to underestimate the true data variability. However, the method that includes fixed effects performs much better. The standard deviation underestimation is only 15 per cent and the average imputed standard deviation is not statistically different from the true population parameter. In sum, these simulation results clearly indicate that there are important pay-offs from including both weights and fixed effects in the imputation routines of the KILM data set. In terms of average unemployment rates, the weighting scheme reduces the bias for about 50 per cent, the remaining 50 per cent reduction is due to the operation of fixed effects. In addition, the estimation of country-level fixed effects also helps to preserve an important proportion of the real variability. 5. Employment distribution by economic activity In this section, the methodology is generalized in order to impute and predict the evolution of employment in different economic sectors. Some empirical evidence suggests a strong correlation between per capita income and the share of GDP of various economic sectors. Similar correlation was found by Schaible and Mahadevan-Vijaya (2002) for a crosssection of countries in the KILM data set. In a very well-known study Fuchs (1980) suggested that the relationship between the importance of the different economic sectors and the per capita income is far from linear. Indeed, when per capita income increases, the share of employment in agriculture exponentially declines; the share of employment in industry 14 initially increases, then reaches a plateau and decreases; and the share of employment in services increases exponentially. In order to determine the best specification in the case of the KILM data set, several semi-parametric regressions between the share of employment in each economic sector and the per capita income were run. The results of this exercise are depicted in Figure 5.1. Figure 5.1: Employment shares and per capita income by economic sector Fractional Polynomial (0 0) .942391 Component+residual for AGRIC Component+residual for INDUS .642 Fractional Polynomial (3 3) -.025791 1.82689 lpcgdp 2.37374 -.05684 1.82689 lpcgdp 2.37374 Fractional Polynomial (-2 3) .824506 Component+residual for SERV .036 1.82689 lpcgdp 2.37374 The top left panel in Figure 5.1 shows the relationship between income and the employment share in agriculture; the top right for the industry; bottom left panel for the service sector. Overall, it is possible to say that these relationships reproduce the pattern found by Fuchs (1980) more than 20 years ago. These findings confirm that per capita GDP is a key predictor of economic structure. In what follows, the methodology applied above is adapted in order to deal with employment distribution by sector and the results of the predictions are described. Table 5.1 summarizes the response rates for employment by sector data across the different subregions. The global response rate is 44 per cent (slightly higher than for unemployment figures). Response rates of over 50 per cent apply to almost every subregion except sub-Saharan Africa, whose response rates are very low. This means we need to collapse this subregion with the Middle East and North Africa into one macro region. Several interpolation techniques were used to boost the response rates by filling the gaps between reported years for the respondent countries. Table 5.2 shows that the response rates increase very little. 15 Table 5.1: Response rates by subregion before interpolation 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Total Major Europe Major Non-Europe Other Europe Eastern Europe Baltic States CIS Melanesia Eastern Asia South-Central Asia South-Eastern Asia Central America South America Eastern Africa Middle Africa Southern Africa Western Africa Middle East North Africa Total 1.00 1.00 0.50 0.33 0.67 0.58 0.67 0.86 0.50 0.55 0.74 0.67 0.19 0.00 0.00 0.00 0.13 0.40 0.48 1.00 1.00 0.50 0.33 0.67 0.58 0.67 0.86 0.50 0.73 0.74 1.00 0.13 0.00 0.00 0.06 0.13 0.40 0.51 1.00 1.00 0.50 0.42 0.67 0.67 0.67 0.86 0.50 0.82 0.74 0.92 0.06 0.00 0.00 0.00 0.13 0.40 0.51 1.00 1.00 0.50 0.58 0.67 0.67 0.33 0.86 0.50 0.64 0.79 0.83 0.19 0.00 0.00 0.00 0.13 0.20 0.51 1.00 1.00 0.50 0.58 0.67 0.67 0.33 0.86 0.63 0.64 0.79 0.92 0.13 0.00 0.20 0.06 0.20 0.40 0.54 1.00 1.00 0.00 0.67 0.67 0.58 0.33 0.86 0.38 0.55 0.79 0.83 0.00 0.00 0.00 0.00 0.20 0.20 0.48 1.00 1.00 0.00 0.67 1.00 0.67 0.33 0.86 0.25 0.64 0.79 0.75 0.00 0.00 0.00 0.00 0.13 0.40 0.49 1.00 1.00 0.00 0.67 1.00 0.67 0.33 0.71 0.25 0.55 0.84 0.83 0.06 0.00 0.20 0.00 0.13 0.40 0.50 1.00 1.00 0.50 0.67 1.00 0.75 0.00 0.71 0.13 0.45 0.68 0.92 0.06 0.00 0.20 0.00 0.20 0.40 0.49 0.95 1.00 1.00 0.67 1.00 0.42 0.00 0.71 0.38 0.36 0.53 0.67 0.00 0.00 0.40 0.00 0.20 0.20 0.43 0.95 1.00 1.00 0.67 1.00 0.33 0.00 0.43 0.00 0.18 0.53 0.75 0.00 0.00 0.00 0.00 0.13 0.00 0.37 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 0.08 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.91 0.92 0.42 0.52 0.75 0.55 0.31 0.71 0.33 0.51 0.68 0.76 0.07 0.00 0.08 0.01 0.14 0.28 0.44 Source: ILO, KILM, 3rd edition. From Table 5.1 it can be concluded that there are missing values in every subregion (or region) and thus it is necessary to impute in every single geographic group. The estimation and imputation phase is similar to that applied for the unemployment rates. First, a missing values mechanism is estimated by running several logit regressions where the dependent variable is 1 if the country reports on distribution of employment by sector and 0 if the report is missing. The explanatory variables are those used in the case of unemployment rates. Several response probabilities are then predicted and country-level weights computed for the respondent countries. The results of this sequence of probit regressions are shown in Table 5.3. In eight out of the ten geographical groups, the assumption that missing values are missing completely at random is demonstrably rejected. Again, the per capita income and the size of the country are two important predictors of response probability. This implies that respondent countries are clearly different from non-respondent ones and that some gains can be obtained by weighting the sample of respondent countries. 16 Table 5.2: Response rates by subregion after interpolation 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Total Major Europe Major Non-Europe Other Europe Eastern Europe Baltic States CIS Melanesia Eastern Asia South-Central Asia South-Eastern Asia Central America South America Eastern Africa Middle Africa Southern Africa Western Africa Middle East North Africa Total 1.00 1.00 0.50 0.33 0.67 0.58 0.67 0.86 0.50 0.55 0.74 0.67 0.19 0.00 0.00 0.00 0.13 0.40 0.48 1.00 1.00 0.50 0.33 0.67 0.58 0.67 0.86 0.50 0.73 0.74 1.00 0.19 0.00 0.00 0.06 0.13 0.40 0.52 1.00 1.00 0.50 0.42 0.67 0.67 0.67 0.86 0.50 0.82 0.79 0.92 0.19 0.00 0.00 0.00 0.13 0.40 0.53 1.00 1.00 0.50 0.58 0.67 0.67 0.33 0.86 0.50 0.73 0.79 0.92 0.25 0.00 0.00 0.00 0.13 0.40 0.53 1.00 1.00 0.50 0.58 0.67 0.67 0.33 0.86 0.63 0.73 0.79 0.92 0.13 0.00 0.20 0.06 0.20 0.40 0.54 1.00 1.00 0.50 0.67 0.67 0.67 0.33 0.86 0.50 0.64 0.79 0.92 0.06 0.00 0.20 0.00 0.27 0.40 0.53 1.00 1.00 0.50 0.67 1.00 0.67 0.33 0.86 0.50 0.64 0.79 0.92 0.06 0.00 0.20 0.00 0.20 0.40 0.53 1.00 1.00 0.50 0.67 1.00 0.67 0.33 0.71 0.50 0.55 0.84 0.92 0.06 0.00 0.20 0.00 0.20 0.40 0.53 1.00 1.00 0.50 0.67 1.00 0.75 0.00 0.71 0.38 0.45 0.79 0.92 0.06 0.00 0.40 0.00 0.27 0.40 0.52 0.95 1.00 1.00 0.67 1.00 0.42 0.00 0.71 0.38 0.36 0.63 0.83 0.00 0.00 0.40 0.00 0.20 0.20 0.46 0.95 1.00 1.00 0.67 1.00 0.33 0.00 0.43 0.00 0.18 0.58 0.75 0.00 0.00 0.00 0.00 0.13 0.00 0.38 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 0.08 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.91 0.92 0.54 0.52 0.75 0.56 0.31 0.71 0.41 0.53 0.70 0.81 0.10 0.00 0.13 0.01 0.17 0.32 0.46 Source: Author s calculations. Table 5.3 Determinants of response probability: Employment distribution by economic sector Developed Europe PCGDP GROWTH SIZE HIPC Constant N Pseudo-R2 LR Chi2 Developed Eastern CIS Eastern Asia Central Asia South-East Central Asia America South America Africa and Middle East Non-Europe Europe 4.513 (3.55)** -0.096 (0.61) 1.282 (4.19)** -20.073 (2.29)* 0.282 (0.84) 2.116 (2.04)* 0.206 (0.27) -0.094 (2.47)* 0.33 (1.40) -49.203 (3.66)** 231 0.46 40.8** 182.056 (2.32)* 60 0.2 6.4 -0.714 (1.15) -0.003 (0.11) 1.158 (3.90)** 1.722 (2.57)* -5.451 -4.566 (0.74) (0.8) 143 132 0.15 26.4* 0.17 30.6** 3.036 (3.62)** 0.226 (1.75) 0.52 (2.22)* 8.269 (3.49)** -29.145 (3.36)** 77 0.59 48.3** 2.16 2.254 2.443 -1.807 1.216 (3.09)** (4.54)** (5.30)** -1.35 (6.13)** -0.211 -0.115 -0.057 0.033 0.014 (1.12) (1.65) (0.94) -0.29 -0.68 0.294 0.916 0.837 1.386 0.705 (2.32)* (5.25)** (4.97)** (3.65)** (6.29)** -0.249 3.881 -4.71 0.642 (0.3) (3.64)** (3.36)** (1.43) -17.711 -24.726 -25.646 4.963 -17.867 (3.14)** (5.09)** (5.42)** -0.46 (7.86)** 80 154 228 132 715 0.15 15.8 0.51 0.32 0.71 109.6** 90.24** 93.7** 0.21 101.6** Notes: Absolute value of the Z statistic is in parenthesis. (*) significant at 5%, (**) significant at 1%. N = No. of observations. 17 After estimating the weights, the final step is to run a sequence of subregional weighted regressions. The prediction model is built on the basis of the empirical evidence summarized in Figure 5.1. That is, for each economic sector the following model for employment shares is estimated: y T Yitk = ln itk 1 yitk ' = + xit + it (17) where yitk is the observed employment participation for economic sector k, country i and period t and xit is a set of covariates explaining this share. In this section this set of covariates includes both the level of the per capita income and its square value. One important difference here is that is no longer country-specific. We proceeded in this way because, on the one hand, the inclusion of a fixed effect did not greatly affect the estimated correlations and, on the other, it did generate some implausible imputations in some subregions. One notable constraint to model (17) is that the add of the employment shares across all the economic sector should always add 1, that is: y k itk = 1 i, t (18) As a consequence, model (17) was estimated as a system of equations for two sectors: agriculture and industry. The default sector is services, whose shares are computed as: yitk = 1 yit ,agr yit ,ind i, t (19) After estimation this model can be used for imputation and prediction. Although it is not strictly necessary to carry out imputations for non-respondent countries, doing so is useful to produce a complete data set that can be used in order to compute additional statistics of labour market conditions or to generate different regional and subregional aggregations. Tables 5.4 and 5.5 show the estimates and projections for the distribution of total employment by sector. One important point is that the model is able to impute shares only; hence, to recover a total, some information about total employment will be needed. This figure is derived from the results on unemployment rates shown in previous sections. A negative side-effect of this approach is that the standard errors reported in Tables 5.4 and 5.5 will clearly be underestimated because they take as true a total employment value that was predicted with some error. In terms of the results, Table 5.5 suggests a reduction in employment, accounted for by the agricultural sector, from 42 per cent in 1996 to 20 per cent by 2015. This is a very significant drop and, as a consequence, agriculture is the only sector where a reduction in the absolute number of workers is expected. This fall for agriculture allows for a marginal increase in the participation of industry (from 22 per cent to 26 per cent) and, more remarkably, for a substantial increase in the service sector (from 37 per cent to 53 per cent). Therefore, although across the entire period it is expected that the total number of workers in industry will increase by about 300 million, the increased employment in services will be two times larger. Appendix tables A.16-A.21 present the same results by subregions. The pattern of change is always similar. Although a decrease in the participation of agriculture is confirmed for every subregion, important differences are observed across subregions. At one extreme are the countries of Eastern Asia where significant structural change is expected and where 18 the share of agriculture will fall from 57 per cent to 13 per cent by 2015, and at the other is the sub-Saharan African region, which is the most stable subregion and where a decrease from 48.5 per cent to just 47.2 per cent is expected. For the manufacturing sector, the patterns vary depending on the initial condition of the countries. We expect a fall in industrial employment in developed countries (Europe and Non-Europe), a rather stable situation in the middle-income countries of Central and South America and an increase in manufacturing employment in the rest of the world. This increase is expected to be relatively significant in South Central Asia (15 per cent to 28 per cent). As a consequence of these compensating trends, the world levels for manufacturing look fairly stable. Finally, all the countries will show growth in the participation of the service sector, particularly in East Asia where it is predicted to rise from 17 per cent in 1996 to 55 per cent by 2015), and also in South-Eastern Asia (37 per cent to 56 per cent). The remaining subregions will also show an increase but movements will be less dramatic. Table 5.4: Employment distribution by economic sector ( 000) Year Agriculture Standard deviation Industry Standard deviation Services Standard deviation 1996 1997 1998 1999 2000 2001 948482.1 47828.1 621052.5 8421.6 2002 920249.8 48151.5 642225.4 8874.9 2003 911250.7 48384.1 661169.6 9398.5 2004 899601.8 48650.6 681875.9 10100.2 2015 651098.2 30328.1 876295.6 18421.8 993981.5 1016692.2 1031457.9 1017261.3 1014201.0 46914.3 566703.4 6845.7 922831.9 47411.2 47111.0 564464.8 6961.9 949289.7 47622.6 47317.5 552916.7 7130.4 47428.2 562826.4 7677.3 47619.6 585472.0 8134.3 983401.2 1028806.1 1058839.8 1132319.4 1176714.3 1211930.9 1250697.2 1752359.0 47851.7 48045.6 48309.4 48563.8 48962.6 49288.4 49688.0 35484.6 Table 5.5: Employment distribution by economic sector (%) Year Agriculture Standard deviation Industry Standard deviation Services Standard deviation 1996 40.0 1.9 22.8 0.3 37.2 1.9 1997 40.2 1.9 22.3 0.3 37.5 1.9 1998 40.2 1.8 21.5 0.3 38.3 1.9 1999 39.0 1.8 21.6 0.3 39.4 1.8 2000 38.1 1.8 22.0 0.3 39.8 1.8 2001 35.1 1.8 23.0 0.3 41.9 1.8 2002 33.6 1.8 23.4 0.3 43.0 1.8 2003 32.7 1.7 23.7 0.3 43.5 1.8 2004 31.8 1.7 24.1 0.4 44.2 1.8 2015 19.9 0.9 26.7 0.6 53.4 1.1 6. Conclusions This report has proposed an integrated methodological approach aimed at generating imputations and projections from the KILM data set, which is a substantial undertaking designed by the ILO to collect and improve the coordination of data on various labour market indicators submitted by its member countries. The team responsible for KILM has concentrated first on widening geographical coverage and improving the timeliness of the information for the core indicators and, second, on expanding the set of indicators to further meet demand for realistic coverage of the world s labour markets. The scope and complexity of the current data set affect the comparability of statistics among countries and regions. 19 Three shortcomings were detected: not all countries report information; many reporting countries report incomplete data; and, finally, not all the reported information is comparable across countries due to different methodological approaches. Improving the KILM data set will require more standardization of the different methodologies involved in data collection and processing than are currently being implemented by member countries. It will also necessitate achieving higher response rates. This entails a prolonged process of continuous investment, the results of which can only be looked for in the long term. Meanwhile, the methodological approach proposed in this investigation is designed to deal with the three shortcomings detected. The issue of incomplete information is overcome by the use of ad hoc country-level interpolation techniques. This also preserves much of the country-level heterogeneity. In addition, modelling the missingness mechanism and weighting the sample of respondent countries deals with the issue of missing countries. The idea of introducing weights is to try to reduce the influence in the sample of those respondent countries that are vastly different. Finally, heterogeneity at country level was controlled for by adopting panel data techniques. The tests achieved from simulations of this approach are encouraging. A remarkable reduction of the bias in comparison with more standard alternatives is observed with 50 per cent of bias reduction generated by the introduction of the weighting scheme used in the model, and the remaining 50 per cent resulting from working with fixed effects. Another significant result of the simulation is that this methodology preserves a large proportion of the original data variability. Of course, research into the consequences and properties of the suggested approach does not end with this investigation. Much still remains to be analysed, in particular how to achieve a more appropriate measurement of uncertainty. The present paper is only a minor step in this direction. A major advance would be to take into account (in the projections) the uncertainties that relate to the estimated fixed effects and, also, the proportion of observations that are imputed as opposed to true data. The method described in this report still treats the different dependent variables (such as unemployment sub-components) as single equations. Thus, all the information related to correlation among them is omitted, leading to an additional loss of information. More structural approaches will be necessary to deal with such issues. These alternative approaches would also require more highly sophisticated estimation techniques (such as maximum likelihood for incomplete data), and are beyond the scope of the present research. However, they must form part of the future research agenda. 20 Bibliograph FUCHS, V. 1980. "Economic growth and the rise of service employment," NBER Working Paper. HOROWITZ, J.; MANSKI, C.1998. "Censoring of outcomes and regressors due to survey nonresponse: Identification and estimation using weights and imputations," Journal of Econometrics, Vol. 84, No. 1, pp. 37-58. KILM. 2003. Key Indicators of the Labour Market, 3rd edition. Geneva, ILO. LITTLE, R., RUBIN, D. 1987: Statistical Analysis with Missing Data. New York, Wiley. SCHAFER, J.L. 1997. Analysis of Incomplete Multivariate Data. London, Chapman and Hall. SCHAIBLE, W. 2000: "Methods for producing world and regional estimates for selected key indicators of the labour market," Employment Sector Paper. Geneva, ILO. Available online at http://www.ilo.org/public/english/employment/strat/publ/ep00-6.htm SCHAIBLE, W.; MAHADEVAN-VIJAYA, R. 2002. "World and regional estimates for selected key indicators of the labour market". Employment Paper No. 2002/36. Available online at http://www.ilo.org/public/english/employment/strat/download/ep36.pdf 21 Appendix Table A.1: Total Male and Female Unemployment ( 000) MFU Europe 1995 20258 0 12052 0 6957 17 CIS 9183 24 21134 1002 25127 916 9307 195 4462 7 9409 1 22706 206 12807 64 1996 20339 0 12182 0 6703 218 9972 11 22475 935 25438 892 9264 26 4042 4 12318 1 23358 235 12219 28 1997 20181 0 11577 0 6179 70 10957 15 23754 875 23421 380 9301 23 3782 4 12653 1 24777 220 12899 67 1998 19158 0 11420 0 6153 26 12350 13 26403 805 24275 427 11426 20 3608 3 14544 1 25482 216 13772 68 1999 18278 0 11286 0 6902 13 13581 18 27442 769 26401 659 12301 24 3142 3 17184 1 26047 181 13740 82 2000 16448 0 10941 0 7615 7 10640 20 25570 833 26027 394 12425 29 3123 5 17107 1 27482 250 14447 36 2001 15354 0 12324 0 7958 6 9768 23 27071 796 25012 520 14335 24 3384 7 15954 1 27991 212 14474 34 2002 16504 0 14287 0 8366 6 9342 15 26151 853 26233 419 16866 42 3465 23 16528 160 29019 296 14798 64 2003 16542 313 14496 328 8262 172 9142 1174 27537 832 27249 416 17178 260 3469 68 16005 244 29663 328 15284 119 2004 16442 312 14009 313 8363 182 9136 1167 27893 843 28898 497 17410 261 3424 67 14658 220 29364 235 15707 122 2015 16428 314 14782 338 7764 174 8190 1085 27164 974 33693 500 19933 276 3967 81 17262 254 37816 282 19097 144 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total 153402 158310 159479 168592 176304 171825 173624 181558 184828 185304 206097 1388 1332 984 939 1033 956 975 1012 1653 1655 1693 Note: The standard deviation is given below unemployment estimates . 22 Table A.2: Total Youth Male Unemployment ( 000) YMU Europe 1995 3129 0 2163 0 1261 5 CIS 1448 6 6090 523 10638 599 2990 136 1256 3 2441 1 8608 113 4240 38 44265 816 1996 3024 0 2203 0 1140 34 1524 5 6283 469 10645 580 3047 14 1129 2 3176 0 8882 129 4050 19 45103 758 1997 2870 0 2133 0 1023 14 1631 6 6417 420 9426 240 3062 12 1009 2 3168 0 9438 129 4254 41 44430 503 1998 2732 0 2095 0 1019 6 1791 7 6815 371 9971 270 3569 11 930 1 3646 1 9778 153 4543 42 46889 485 1999 2657 0 2059 0 1160 3 1977 10 6907 342 10950 418 4192 13 748 1 4239 0 10084 134 4409 51 49382 559 2000 2328 0 1966 0 1221 1 1542 10 6379 365 10625 246 4112 15 809 2 4117 1 10604 179 4677 24 48380 476 2001 2285 0 2170 0 1295 2 1482 11 6687 344 10067 324 4717 12 857 3 3979 1 10727 154 4650 20 48917 498 2002 2510 0 2382 0 1341 2 1367 8 6455 370 10637 263 5427 22 882 6 4096 76 11143 187 4708 30 50948 499 2003 2510 68 2422 94 1144 54 1174 263 6809 362 11111 261 5489 180 875 32 3982 102 11444 219 4845 67 51805 617 2004 2477 67 2352 92 1135 55 1185 263 6954 370 11901 313 5518 180 853 30 3599 94 11391 137 4951 67 52315 621 2015 2351 65 2472 101 821 38 821 162 6264 383 13135 299 5733 181 900 32 3791 94 14845 178 5191 61 56323 597 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Sharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 23 Table A.3: Total Adult Male Unemployment ( 000) AMU Europe 1995 7518 0 4586 0 2429 14 CIS 3534 17 7055 696 4819 452 1885 37 1354 3 2473 0 5557 122 4290 44 45500 841 1996 7739 0 4614 0 2384 186 3877 7 7747 666 4931 445 1873 12 1179 2 3250 0 5685 142 4031 16 47309 835 1997 7670 0 4275 0 2169 60 4254 10 8435 637 4638 193 1924 11 998 2 3231 0 5970 127 4253 44 47815 682 1998 7179 0 4287 0 2219 22 4772 8 9832 598 4816 218 2547 10 977 1 3663 1 5944 95 4411 43 50647 645 1999 6836 0 4271 0 2554 11 5268 11 10415 580 5202 338 2584 12 895 1 4514 0 5961 82 4532 51 53032 678 2000 6117 0 4156 0 2794 6 3961 13 9722 631 5179 206 2690 15 874 2 4555 1 6488 113 4695 20 51229 674 2001 5873 0 4845 0 2928 5 3660 15 10383 607 5129 278 2973 12 951 3 4223 1 6524 86 4700 18 52188 673 2002 6560 0 5765 0 3101 5 3501 10 9985 648 5371 222 3589 23 982 5 4256 57 6731 152 4827 48 54670 706 2003 6599 227 5843 257 3156 122 3495 821 10463 632 5556 220 3694 78 985 42 4066 137 6805 156 5003 83 55666 1146 2004 6581 226 5633 243 3221 129 3487 816 10549 639 5840 260 3786 80 973 41 3610 118 6783 130 5170 86 55633 1147 2015 6858 232 5958 263 3264 129 3480 792 11033 767 7367 282 4897 104 1240 53 4739 147 8869 154 7269 115 64974 1227 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 24 Table A.4: Total Youth Female Unemployment ( 000) YFU Europe 1995 2730 0 1736 0 999 3 CIS 1325 5 4235 339 5421 416 2657 129 953 4 2206 0 5254 88 2386 22 29902 560 1996 2619 0 1740 0 935 22 1396 4 4302 300 5362 399 2625 15 860 2 2932 0 5407 94 2343 10 30519 509 1997 2575 0 1697 0 858 8 1464 6 4365 267 5101 178 2534 13 850 2 3021 0 5795 87 2529 26 30788 334 1998 2350 0 1635 0 816 3 1579 4 4552 233 5115 193 3057 11 820 2 3592 0 6069 107 2787 27 32371 322 1999 2234 0 1611 0 908 2 1620 6 4582 213 5517 288 3310 12 712 2 3995 0 6232 80 2798 34 33519 369 2000 1975 0 1527 0 947 1 1482 7 4255 227 5762 175 3331 13 681 3 4003 0 6541 117 2924 14 33428 310 2001 1772 0 1623 0 1006 1 1352 8 4414 212 5496 232 3908 11 779 4 3823 0 6622 105 2939 14 33734 332 2002 1760 0 1802 0 1064 1 1325 5 4285 229 5636 182 4453 16 796 15 3937 91 6894 134 2976 17 34927 336 2003 1742 44 1828 60 1061 34 1335 273 4483 222 5826 181 4500 155 799 26 3817 107 7123 154 3069 40 35583 474 2004 1717 43 1802 60 1057 36 1340 274 4563 226 6090 210 4524 155 790 26 3558 102 6940 98 3141 41 35522 473 2015 1559 39 1886 65 755 24 875 166 4050 231 6798 201 4686 155 824 28 3748 100 8951 116 3248 37 37379 421 Major Non Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 25 Table A.5: Total Adult Female Unemployment ( 000) AFY Europe 1995 6882 0 3567 0 2267 8 CIS 2876 15 3753 362 4250 319 1775 37 899 4 2289 0 3287 85 1892 13 33735 492 1996 6958 0 3625 0 2245 105 3175 5 4142 348 4500 319 1720 11 875 2 2960 0 3384 97 1796 7 35379 494 1997 7066 0 3472 0 2129 33 3609 7 4537 334 4255 131 1781 10 925 2 3233 0 3575 90 1863 15 36446 372 1998 6896 0 3404 0 2100 12 4209 6 5204 314 4374 158 2252 9 882 2 3644 0 3690 52 2031 14 38685 356 1999 6551 0 3345 0 2281 6 4715 8 5539 306 4731 252 2215 11 787 2 4436 0 3771 44 2001 17 40371 399 2000 6028 0 3292 0 2654 3 3655 9 5214 333 4462 149 2293 14 759 3 4432 1 3850 64 2150 9 38789 371 2001 5424 0 3685 0 2730 3 3274 11 5587 321 4320 184 2736 12 797 4 3929 0 4118 50 2186 16 38785 374 2002 5673 0 4338 0 2861 3 3149 7 5425 345 4589 154 3398 22 805 16 4239 90 4251 107 2286 26 41013 404 2003 5691 200 4403 171 2902 103 3138 748 5782 335 4756 155 3495 71 810 35 4140 136 4291 108 2367 36 41775 902 2004 5668 201 4222 164 2950 110 3123 742 5827 338 5067 192 3583 73 808 35 3891 124 4250 98 2446 38 41834 902 2015 5660 198 4466 176 2923 107 3014 704 5818 401 6393 201 4618 94 1004 44 4985 155 5152 103 3388 50 47420 909 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 26 Table A.6: Male - Female Unemployment Rates (%) MFUR Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 10.1 0 5.2 0 11.6 0 CIS 7 0 2.8 0.1 5.4 0.2 4.2 0.1 7.3 0 6.6 0 10.3 0.1 13.3 0.1 5.90 0.10 10.1 0 5.2 0 11.3 0.4 7.6 0 2.9 0.1 5.3 0.2 4 0 6.5 0 8.6 0 10.3 0.1 12.3 0 6.00 0.10 10 0 4.9 0 10.4 0.1 8.4 0 3 0.1 4.8 0.1 4 0 5.8 0 8.5 0 10.6 0.1 12.5 0.1 5.90 0.00 9.4 0 4.8 0 10.4 0 9.4 0 3.3 0.1 4.8 0.1 4.8 0 5.4 0 9.5 0 10.6 0.1 12.9 0.1 6.20 0.00 8.9 0 4.7 0 11.7 0 10.6 0 3.4 0.1 5.1 0.1 5 0 4.6 0 10.9 0 10.5 0.1 12.3 0.1 6.30 0.00 7.9 0 4.5 0 12.9 0 8.3 0 3.1 0.1 4.9 0.1 4.9 0 4.5 0 10.7 0 10.8 0.1 12.6 0 6.10 0.00 7.4 0 5.1 0 13.3 0 7.6 0 3.3 0.1 4.6 0.1 5.5 0 4.8 0 9.8 0 10.7 0.1 12.3 0 6.00 0.00 7.9 0 5.8 0 14 0 7.3 0 3.1 0.1 4.8 0.1 6.3 0 4.9 0 9.9 0.1 10.8 0.1 12.2 0.1 6.20 0.00 7.9 0.1 5.9 0.1 13.8 0.3 7.1 0.9 3.3 0.1 4.8 0.1 6.3 0.1 4.8 0.1 9.4 0.1 10.8 0.1 12.2 0.1 6.20 0.10 7.8 0.1 5.6 0.1 13.9 0.3 7 0.9 3.3 0.1 5 0.1 6.3 0.1 4.6 0.1 8.5 0.1 10.4 0.1 12.2 0.1 6.10 0.10 7.7 0.1 5.7 0.1 13.2 0.3 6.2 0.8 3 0.1 4.6 0.1 5.9 0.1 4.3 0.1 8.4 0.1 10.2 0.1 11.3 0.1 5.90 0.00 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 27 Table A.7: Youth Male Unemployment Rates (%) YMUR Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 17.8 0 11.4 0 24.4 0.1 CIS 13.9 0.1 6.6 0.6 14.4 0.8 9.4 0.4 10.9 0 10.9 0 22.2 0.3 24.8 0.2 13.0 0.2 17.5 0 11.7 0 23.1 0.7 15.1 0.1 7 0.5 14.2 0.8 9.3 0 9.7 0 14.4 0 22.2 0.3 23.1 0.1 13.3 0.2 16.9 0 11.3 0 21.3 0.3 16 0.1 7.4 0.5 12.4 0.3 9.4 0 8.7 0 14 0 22.8 0.3 23.4 0.2 13.1 0.1 16 0 11.1 0 21.8 0.1 17.3 0.1 8.1 0.4 12.9 0.3 11.4 0 7.9 0 15.9 0 22.8 0.4 24.1 0.2 13.8 0.1 15.8 0 10.9 0 25.4 0.1 19.2 0.1 8.4 0.4 13.9 0.5 12.4 0 6.5 0 18.1 0 22.8 0.3 22.2 0.3 14.3 0.2 14 0 10.4 0 27.6 0 16.2 0.1 7.9 0.4 13.2 0.3 12.1 0 7 0 17.6 0 23.1 0.4 23 0.1 14.0 0.1 14 0 11.6 0 29.5 0 15.3 0.1 8.2 0.4 12.3 0.4 13.6 0 7.5 0 17.1 0 22.7 0.3 22.6 0.1 14.0 0.1 15.7 0 12.7 0 31.5 0 14.5 0.1 7.8 0.5 12.7 0.3 15.5 0.1 7.8 0.1 17.7 0.3 22.8 0.4 22.7 0.1 14.4 0.1 15.7 0.4 12.8 0.5 27.3 1.3 12.3 2.8 8.1 0.4 13 0.3 15.5 0.5 7.6 0.3 17.1 0.4 22.7 0.4 22.7 0.3 14.4 0.2 15.5 0.4 12.4 0.5 27.5 1.3 12.2 2.7 8.1 0.4 13.6 0.4 15.4 0.5 7.4 0.3 15.4 0.4 22 0.3 22.7 0.3 14.3 0.2 15.3 0.4 12.5 0.5 27.3 1.3 11 2.2 7.6 0.5 12.9 0.3 15.4 0.5 7.2 0.3 16.2 0.4 22.2 0.3 22.4 0.3 14.4 0.2 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 28 Table A.8: Adult Male Unemployment Rates (%) AMUR Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 7.4 0 4.2 0 8.7 0 CIS 6 0 2.1 0.2 1.9 0.2 1.9 0 4.7 0 3.8 0 6.4 0.1 7.6 0.1 3.7 0.1 7.6 0 4.1 0 8.6 0.7 6.6 0 2.3 0.2 1.9 0.2 1.9 0 3.9 0 4.9 0 6.4 0.2 6.9 0 3.8 0.1 7.5 0 3.8 0 7.8 0.2 7.3 0 2.5 0.2 1.7 0.1 1.9 0 3.2 0 4.8 0 6.5 0.1 7.1 0.1 3.8 0.1 6.9 0 3.8 0 8 0.1 8.2 0 2.8 0.2 1.8 0.1 2.4 0 3.1 0 5.3 0 6.4 0.1 7.2 0.1 3.9 0.1 6.6 0 3.7 0 9.2 0 9.3 0 2.9 0.2 1.9 0.1 2.4 0 2.8 0 6.4 0 6.2 0.1 7.1 0.1 4.0 0.1 5.8 0 3.6 0 10.1 0 7 0 2.7 0.2 1.8 0.1 2.4 0 2.6 0 6.3 0 6.6 0.1 7.2 0 3.8 0.1 5.6 0 4.2 0 10.4 0 6.4 0 2.8 0.2 1.8 0.1 2.6 0 2.8 0 5.7 0 6.5 0.1 7 0 3.8 0.0 6.2 0 4.9 0 11 0 6.1 0 2.7 0.2 1.8 0.1 3.1 0 2.8 0 5.7 0.1 6.5 0.1 7 0.1 4.0 0.1 6.2 0.2 4.9 0.2 11.1 0.4 6.1 1.4 2.8 0.2 1.8 0.1 3.1 0.1 2.7 0.1 5.3 0.2 6.5 0.1 7 0.1 4.0 0.1 6.1 0.2 4.7 0.2 11.3 0.5 6 1.4 2.8 0.2 1.9 0.1 3.1 0.1 2.6 0.1 4.6 0.1 6.3 0.1 7 0.1 3.9 0.1 6.2 0.2 4.8 0.2 11.2 0.4 5.6 1.3 2.6 0.2 1.8 0.1 3.1 0.1 2.6 0.1 4.8 0.1 6.1 0.1 7.1 0.1 3.8 0.1 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 29 Table A.9: Youth Female Unemployment Rates (%) YFUR Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 20.5 0 10.4 0 25.7 0.1 CIS 14.3 0 4.8 0.4 15.7 1.2 10.9 0.5 16.1 0.1 15.5 0 17.3 0.3 34.5 0.3 12.1 0.2 20.3 0 10.5 0 25.4 0.6 15.4 0 5 0.4 15.5 1.2 10.8 0.1 14.5 0 20.4 0 17.2 0.3 32 0.1 12.4 0.2 20.6 0 10.2 0 23.7 0.2 16.3 0.1 5.3 0.3 14.6 0.5 10.6 0.1 13.9 0 20.4 0 17.8 0.3 32.7 0.3 12.6 0.1 18.8 0 9.7 0 23 0.1 17.7 0 5.7 0.3 14.5 0.5 12.8 0 13 0 23.5 0 18 0.3 34 0.3 13.3 0.1 17.8 0 9.6 0 25.9 0 19.6 0.1 5.9 0.3 15.6 0.8 13.8 0 11.4 0 25.1 0 17.9 0.2 31.4 0.4 13.8 0.2 16 0 9 0 27.6 0 17.5 0.1 5.6 0.3 16 0.5 13.5 0.1 10.9 0 25 0 18.1 0.3 32.9 0.2 13.6 0.1 14.8 0 9.7 0 29.6 0 15.9 0.1 5.8 0.3 15.4 0.7 15.6 0 12.6 0.1 24 0 17.7 0.3 32.4 0.2 13.7 0.1 15.2 0 10.7 0 32 0 15.3 0.1 5.6 0.3 15.7 0.5 17.6 0.1 12.9 0.2 24.8 0.6 17.9 0.3 32.5 0.2 14.1 0.1 15 0.4 10.8 0.4 32.4 1.1 15.2 3.1 5.8 0.3 15.9 0.5 17.6 0.6 12.8 0.4 23.9 0.7 18 0.4 32.6 0.4 14.1 0.2 14.8 0.4 10.5 0.3 32.9 1.1 15.1 3.1 5.8 0.3 16.3 0.6 17.6 0.6 12.6 0.4 22.1 0.6 17 0.2 32.6 0.4 13.9 0.2 14.2 0.4 10.7 0.4 32.5 1 13.2 2.5 5.5 0.3 15.9 0.5 17.7 0.6 12.4 0.4 23 0.6 17.1 0.2 32 0.4 14.0 0.2 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment/employment estimates 30 Table A.10: Adult Female Unemployment Rates (%) AFUR Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 10 0 4.2 0 9.7 0 CIS 5.4 0 1.5 0.1 4.1 0.3 2.5 0.1 6.2 0 5.6 0 5.1 0.1 11.9 0.1 4.2 0.1 10 0 4.2 0 9.7 0.5 6 0 1.6 0.1 4.1 0.3 2.3 0 5.8 0 7.2 0 5.1 0.1 10.7 0 4.3 0.1 10.1 0 4 0 9.2 0.1 6.8 0 1.7 0.1 3.8 0.1 2.4 0 5.7 0 7.4 0 5.2 0.1 10.6 0.1 4.3 0.0 9.7 0 3.8 0 9.1 0.1 7.8 0 1.9 0.1 3.7 0.1 2.9 0 5.2 0 8 0 5.2 0.1 11 0.1 4.5 0.0 9 0 3.7 0 9.8 0 9 0 1.9 0.1 3.9 0.2 2.8 0 4.6 0 9.2 0 5.2 0.1 10.3 0.1 4.6 0.0 8.2 0 3.6 0 11.3 0 6.9 0 1.8 0.1 3.5 0.1 2.7 0 4.3 0 9.1 0 5.2 0.1 10.7 0 4.3 0.0 7.3 0 4 0 11.5 0 6.1 0 1.9 0.1 3.4 0.1 3.2 0 4.4 0 7.8 0 5.4 0.1 10.3 0.1 4.2 0.0 7.6 0 4.7 0 11.9 0 5.9 0 1.8 0.1 3.5 0.1 3.9 0 4.3 0.1 8.2 0.2 5.4 0.1 10.4 0.1 4.4 0.0 7.6 0.3 4.7 0.2 12 0.4 5.8 1.4 1.9 0.1 3.5 0.1 3.9 0.1 4.2 0.2 7.8 0.3 5.4 0.1 10.4 0.2 4.4 0.1 7.5 0.3 4.5 0.2 12.2 0.5 5.8 1.4 1.9 0.1 3.7 0.1 3.9 0.1 4.1 0.2 7.1 0.2 5.2 0.1 10.4 0.2 4.3 0.1 7.5 0.3 4.5 0.2 12 0.4 5.4 1.3 1.7 0.1 3.6 0.1 3.9 0.1 4 0.2 7.4 0.2 4.8 0.1 10.3 0.2 4.2 0.1 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below unemployment estimates 31 Table A.11: Male - Female Employment-to-Population (%) MFUPOP Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 50.4 0 61.3 0 52.5 0 CIS 57.1 0 78 0.1 56.6 0.1 67.8 0.1 55.9 0 61.5 0 66 0.1 45 0 63.0 0.0 50.1 0 61.4 0 51.8 0.2 56 0 77.9 0.1 56.7 0.1 68.2 0 56.5 0 59.6 0 66.1 0.1 45.5 0 62.8 0.0 50 0 61.8 0 51.8 0.1 55.3 0 77.8 0.1 57.2 0 67.8 0 57.7 0 60.4 0 65.9 0.1 45.6 0 62.9 0.0 50.6 0 61.8 0 51.3 0 54.4 0 77.4 0.1 57.2 0.1 66.8 0 58.1 0 60 0 66 0.1 45.6 0 62.7 0.0 51 0 61.8 0 50.4 0 52.2 0 77.3 0.1 57.1 0.1 67.2 0 58.1 0 59.9 0 66.1 0.1 46.4 0 62.6 0.0 51.5 0 61.8 0 49.4 0 53.1 0 77.2 0.1 57.2 0 67.7 0 58 0 59.5 0 66.1 0.1 46 0 62.7 0.0 51.6 0 61.2 0 49.3 0 53.4 0 77 0.1 57.2 0.1 67.5 0 57.5 0 60.2 0 66.1 0.1 46 0 62.6 0.0 51.2 0 60.6 0 48.8 0 53.4 0 76.9 0.1 57 0 67 0 57.3 0 59.8 0.1 66 0.1 45.9 0 62.4 0.0 51 0.1 60.4 0.1 48.8 0.2 53.4 0.5 76.6 0.1 57 0 67.1 0.1 57.4 0.1 60.2 0.1 66.1 0.1 45.9 0.1 62.3 0.0 51 0.1 60.4 0.1 48.7 0.2 53.4 0.5 76.4 0.1 56.8 0.1 67.2 0.1 57.5 0.1 60.8 0.1 66.3 0.1 46 0 62.3 0.0 48.9 0.1 58.1 0.1 48.3 0.2 54.2 0.5 74.3 0.1 57.2 0 67.8 0.1 57.7 0.1 60.5 0.1 67 0.1 47.4 0 61.7 0.0 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below employment estimates 32 Table A.12: Youth Male Employment-to-Population (%) YMUPOP Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 44.1 0 52.8 0 37.9 0 CIS 41 0 71.7 0.4 52.2 0.5 58.2 0.3 62.4 0 63.8 0 56.1 0.2 41.1 0.1 56.8 0.2 44 0 52.3 0 36.6 0.3 38.6 0 71.1 0.4 52.1 0.5 59 0 62.8 0 59.6 0 56.2 0.2 41.6 0.1 56.2 0.1 44 0 52.4 0 36.3 0.1 38.2 0 70.4 0.4 53 0.2 58.1 0 63.3 0 60.1 0 55.8 0.2 41.5 0.1 56.0 0.1 45.1 0 52.4 0 35 0.1 37.7 0 69.2 0.3 52.4 0.2 53.9 0 63.8 0 58.4 0 55.9 0.3 41.3 0.1 55.1 0.1 44.8 0 52.4 0 32.7 0 36.2 0 68.4 0.3 51.7 0.3 56.7 0 63.6 0 57.1 0 56 0.2 43 0.1 54.9 0.1 45.5 0 53 0 30.8 0 34.3 0 68.3 0.3 52.3 0.2 56.4 0 62.4 0 56.6 0 56.2 0.3 42.2 0.1 54.7 0.1 44.6 0 51.3 0 29.9 0 34.6 0 67.5 0.3 52.6 0.2 56.1 0 60.9 0 56 0 56.4 0.2 41.6 0.1 54.3 0.1 43.2 0 50.4 0 28.5 0 33.4 0 67.4 0.3 52.3 0.2 54.9 0 59.8 0 54.8 0.2 56.3 0.3 41 0.1 53.8 0.1 43.1 0.2 50.5 0.3 30.1 0.5 34.3 1.1 67 0.3 52.2 0.2 54.9 0.3 59.9 0.2 55.2 0.3 56.4 0.3 41.1 0.2 53.8 0.1 43.2 0.2 50.9 0.3 29.9 0.5 34.5 1.1 67 0.3 51.9 0.2 55 0.3 60.1 0.2 56.4 0.3 57 0.2 41.2 0.2 54.0 0.1 43.3 0.2 51.6 0.3 29.7 0.5 37.4 0.9 69 0.3 52.8 0.2 55.2 0.3 60.7 0.2 54.9 0.3 57.4 0.2 42.1 0.1 55.0 0.1 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below employment estimates 33 Table A.13: Adult Male Employment-to-Population (%) AMUPOP Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 66.5 0 75 0 66.1 0 CIS 71.8 0 87.7 0.2 88.6 0.2 89.3 0 84.1 0 83.8 0 87.5 0.1 82.1 0.1 82.6 0.1 65.9 0 75 0 65.6 0.5 70.7 0 87.4 0.2 88.6 0.2 89.2 0 84.6 0 82.1 0 87.4 0.1 82.4 0 82.4 0.1 65.8 0 75.2 0 65.6 0.2 69.4 0 87.1 0.2 88.7 0.1 89 0 85.7 0 82.5 0 87.1 0.1 82.4 0.1 82.3 0.0 66.2 0 75 0 65 0.1 68 0 86.7 0.2 88.6 0.1 88.4 0 85.7 0 82 0 87.2 0.1 82.4 0.1 82.1 0.0 66.1 0 74.8 0 63.9 0 65.6 0 86.4 0.1 88.4 0.1 88.2 0 85.6 0 81.1 0 87.2 0.1 82.6 0.1 81.8 0.0 66.5 0 74.6 0 62.8 0 66.7 0 86.5 0.2 88.4 0.1 88.2 0 85.4 0 80.8 0 86.8 0.1 82.3 0 81.9 0.0 66.4 0 73.8 0 62.8 0 67.2 0 86.2 0.1 88.4 0.1 88 0 85 0 81.3 0 86.8 0.1 82.3 0 81.8 0.0 65.8 0 72.9 0 62.2 0 67.3 0 86.2 0.2 88.3 0.1 87.5 0 84.7 0 81.1 0.1 86.7 0.1 82 0.1 81.6 0.0 65.7 0.1 72.7 0.2 62 0.3 67.3 1 85.9 0.1 88.3 0.1 87.5 0.1 84.8 0.1 81.4 0.2 86.7 0.1 82 0.1 81.5 0.1 65.5 0.1 72.7 0.2 61.8 0.3 67.3 1 85.8 0.1 88.2 0.1 87.4 0.1 84.9 0.1 81.9 0.1 86.9 0.1 82 0.1 81.5 0.1 62.7 0.1 69.1 0.2 59.7 0.3 66.4 0.9 83.3 0.2 87.7 0.1 86.6 0.1 84.2 0.1 80.5 0.1 87 0.1 81.6 0.1 80.5 0.1 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below employment estimates 34 Table A.14: Youth Female Employment-to-Population (%) YFUPOP Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 33.7 0 48.8 0 29.2 0 CIS 37.3 0 74 0.3 26.1 0.4 44.8 0.3 30.1 0 39.1 0 46.3 0.2 15 0.1 43.7 0.1 33 0 48.8 0 27.6 0.2 35.6 0 73.4 0.3 25.7 0.4 44.1 0 30.3 0 36.4 0 46.6 0.2 16 0 42.9 0.1 32.2 0 49.2 0 27.7 0.1 34.5 0 72.7 0.2 25.6 0.2 42.7 0 31.3 0 37 0 46.5 0.2 16.1 0.1 42.2 0.1 33.2 0 49.9 0 27.4 0 33.2 0 71.6 0.2 25.4 0.2 41.2 0 32.3 0 36 0 46.6 0.2 16.2 0.1 41.5 0.1 33.9 0 49.7 0 26 0 29.6 0 70.8 0.2 24.6 0.2 40.4 0 32.7 0 36.2 0 46.8 0.1 17.7 0.1 40.9 0.1 34.2 0 50.2 0 24.9 0 30.7 0 70.4 0.2 24.4 0.1 41.1 0 32.7 0 36.1 0 47.1 0.2 16.7 0 40.7 0.1 33.7 0 49.4 0 24.2 0 30.9 0 69.6 0.2 23.8 0.2 40.4 0 31.5 0 36 0 47.4 0.2 16.7 0 40.1 0.1 32.8 0 48.7 0 23.1 0 31 0 69.4 0.2 23.3 0.1 39.4 0 31 0.1 35.2 0.3 47.3 0.2 16.5 0 39.7 0.1 32.8 0.1 48.8 0.2 22.9 0.4 31.1 1.1 69.1 0.2 23.2 0.1 39.5 0.3 31 0.2 35.7 0.3 47.3 0.2 16.5 0.1 39.7 0.1 32.9 0.1 49.1 0.2 22.6 0.4 31.2 1.1 69.1 0.2 23.1 0.2 39.5 0.3 31.2 0.1 36.6 0.3 47.9 0.1 16.5 0.1 39.8 0.1 32.8 0.1 49.5 0.2 22.4 0.3 33.3 1 70.6 0.2 22.9 0.1 39.4 0.3 31.5 0.2 36.4 0.3 48.3 0.1 16.8 0.1 39.5 0.1 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below employment estimates 35 Table A.15: Adult Female Employment-to-Population (%) AFUPOP Europe 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 40.2 0.1 53 0.1 49.1 0.3 CIS 53.2 0.4 71.5 0.1 37.8 0.1 61.3 0 38.6 0.1 48.4 0.1 61.7 0.1 23.1 0.1 52.6 0 40.2 0.1 53.2 0.1 48.7 0.4 52.7 0.7 71.7 0.1 38.3 0.1 62.3 0.1 39.4 0.1 47.4 0.1 62.1 0.1 23.8 0.1 52.9 0.1 40.1 0.1 53.8 0.1 48.6 0.3 52.5 0.7 72 0.1 39 0 62.3 0.1 41 0.1 48.8 0.1 62.2 0.1 24.2 0.1 53.3 0.1 40.8 0.1 53.8 0.1 48.3 0.2 52.2 0.7 71.9 0.1 39.6 0.1 62.1 0.1 41.1 0.1 49.1 0.1 62.5 0.1 24.5 0.1 53.4 0.1 41.5 0.1 53.9 0.1 48 0.2 50.2 0.8 72.1 0.1 40 0.1 62.3 0.1 41.1 0.1 50 0.1 62.6 0.1 25.3 0.1 53.6 0.1 42 0.1 53.9 0.1 47.2 0.2 51.8 0.9 71.9 0.1 40.2 0 63.4 0.1 41.4 0.1 49.4 0.1 62.7 0.1 25.2 0 53.8 0.1 42.5 0.1 53.7 0.1 47.2 0.2 52 0.8 72 0.1 40.1 0.1 63.3 0.1 41.3 0.1 50.9 0.1 62.7 0.1 25.6 0 53.9 0.1 42.4 0.1 53.3 0.1 47 0.2 52.3 0.8 71.9 0.1 40 0 62.9 0.1 41.3 0.1 50.7 0.1 62.7 0.1 25.9 0 53.8 0.1 42.3 0.1 53.1 0.1 46.8 0.2 52.2 0.8 71.6 0.1 39.9 0 62.9 0.1 41.3 0.1 50.9 0.1 62.7 0.1 25.9 0 53.7 0.1 42.1 0.1 53.1 0.1 46.5 0.2 52.1 0.8 71.3 0.1 39.9 0.1 62.9 0.1 41.3 0.1 51.3 0.1 62.8 0.1 25.9 0 53.5 0.1 39.6 0.1 50.4 0.1 44.5 0.2 50.2 0.7 67.2 0.1 39.4 0 62.2 0.1 40.4 0.1 50.3 0.1 63.3 0.1 25.6 0 51.4 0 Major Non-Europe Eastern Europe + Baltic Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Total Note: The standard deviation is given below employment estimates 36 Table A.16: Total Employment by Economic Sectors: Agriculture ( 000) Region Europe 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 16532.6 15839.7 16075.6 15997.2 14458.5 13672.6 15143.2 14974.4 14645.8 10541.5 773.6 735.1 782.2 814.2 772.2 833.2 778.7 771.4 763.7 516 8435.2 8312.2 8221.8 8065.9 180.8 191 199.7 228.9 11464.4 11411.8 11149.9 159.7 171.3 178.7 7987 272.7 7631 6998.2 6893.9 6686.8 4789.4 261.6 275.5 290.9 327.6 912.6 10689 7335.2 155.1 147.8 7029 6662.3 5514.7 145.1 144.9 387.1 Major Non-Europe Eastern Europe + Baltic 10830 10719.8 175.9 165.7 CIS 32886.8 28688.7 604.6 598.2 28363 24099.1 28146.1 27702.5 29065.6 590.8 585.1 648 712.7 779 28740 28457.5 29908.1 862.4 952.4 1263.6 Eastern Asia 429231 433246. 450653. 456744. 462277. 387689. 377874. 368714. 358524. 113949. 5 2 3 8 2 6 2 5 7 46169 46326.3 46433.9 46447.7 46579.4 46733 47004.3 47171.9 47373 27195.6 South Central Asia 242593. 251332. 255309. 247944. 248535. 254265. 253222. 251557. 247573. 218973. 9 6 9 6 9 8 5 6 8 5 5954.1 6177.9 6468.1 7032.6 7372.3 7672.5 7989.4 8330.5 8677.7 11058.6 South-Eastern Asia 97891 108626. 103508. 91142.8 81971.6 70260.1 65786.7 66147.7 66216 56012.6 3 6 4455.9 4499.6 5017.5 5102.7 5128.2 5121.4 5099.6 5112.5 5095.1 4625.7 14378.3 377.7 15613 14759.9 14869.2 14835.4 14453.1 10780.9 10458.7 10614.3 11565.6 393.4 402.8 406.2 403.2 414.4 426.9 439.1 446.9 509.3 Central America South America 20421.2 20805.4 20472 21767.8 10601.1 23846.2 12710.3 12270.3 12511.3 12185.2 961.6 983.9 1004.4 1024.5 1014.2 1050.9 1069.7 1083.9 1090.7 1017.1 98780 101769. 101374. 103178 109564. 112422. 114903. 117575. 120356. 157689. 6 9 8 7 5 2 3 3 3326.3 3439.7 3548.6 3624.6 3732.5 3840 3938.6 4031.8 4140 5542.3 21367.1 21046.4 21569.1 22622.4 25102.9 25850 26429.2 26889.8 27353.1 29968.7 960.2 990.9 1014.9 1052 1049 1065.6 1086.1 1107.8 1129.3 1282.6 Sub-Saharan Africa Middle East Note: The standard deviation is given below employment estimates 37 Table A.17: Total Employment by Economic Sectors: Industry ( 000) Region Europe 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 52683.8 52560.2 53426.1 247.7 242.5 251.8 53250 54070. 53881.3 8 257.4 257.1 263.6 53441 53649.8 53844.8 53218.6 256.7 257.9 260.3 293.9 Major Non-Europe 58478.3 59391.4 58853.7 58269.6 58437. 57031.1 54949.3 54964.6 54577.5 41121.7 7 1271.1 1364.6 1423.5 1758.5 2258.7 2134.9 2295.2 2465.3 2878.5 7990.5 Eastern Europe + Baltic 17568.2 17502.3 17348.2 16732.6 16200 113 CIS 117.3 121 124.5 126 16285 18116.1 18375.8 18610.4 20427.1 124.8 124.6 129.2 137.4 343.7 39101.5 34937.3 33678.5 31035 37354. 38521.2 38109.2 39108.7 39960.8 41030.9 2 1342.6 1323.9 1257.5 1355.3 1595.5 1821.6 2046.5 2319.3 2607.1 3539.8 Eastern Asia 196576. 195001. 180103. 178791. 18191 213300. 220420. 225529. 230977. 286515. 1 5 9 8 4 4 8 5 7 2 5371.3 5393.6 5414.3 5434.4 5478.6 5531.8 5616.4 5700.5 5808.5 8239.8 69832 71350.4 74776.2 84591.9 91679. 93537.8 99623.7 106734. 115267. 198881. 4 3 1 9 3546.6 3699.1 3989 4725.1 5116.6 5482.3 5937.7 6458.2 7094.8 13724.8 South Central Asia South-Eastern Asia 41305.5 39709.7 36574.7 41929.7 42620. 46912 49900.5 51704.1 53702.7 83214.1 7 1031.5 1033.2 970.1 1015.3 1067.2 1077.6 1110.2 1147.7 1188.3 1904.9 Central America 13150.9 13576.1 15481.3 15283.4 16096. 15960.3 15060.8 15451.5 15824.5 7 198.1 212.8 221.4 230.5 244.3 242.4 243.3 249.5 258.5 29386.1 30521.5 30421.7 29552.1 33341. 30287.5 35465.7 36477.4 1 382.4 409.9 418.9 412.2 413 409.1 403.2 413.5 26685 28305.1 28152.7 29277. 30182.9 3 709.7 730.4 755.2 777.6 795.6 19129 358 South America 37614 46637.3 423.9 677.9 Sub-Saharan Africa 26143.7 691.4 Middle East 31049 31985.7 33195.2 43751.1 816.3 839 864.5 1164.6 22477.4 23229.3 23947.3 25237.7 24480. 1 463.1 471.2 481 497.3 518 25153 26089.3 27188.3 28301.2 42368.6 535.2 553 577.6 603.3 970.1 Note: The standard deviation is given below employment estimates 38 Table A.18: Total Employment by Economic Sectors: Services ( 000) Region Europe 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 111918.7 113620.9 116134.3 812.3 774.1 821.7 118912 122711.9 125244.4 123851.8 124591.2 125562.3 132281.9 853.9 813.9 873.9 819.9 813.4 806.8 593.8 166715 169094.1 170791.1 2150.8 2311.7 2482.4 173604 200200.8 2897.1 8042.4 Major NonEurope 153695.8 156728.8 159644.9 162603.4 165034.4 1283.9 1377.9 1437.4 1773.3 2275.1 Eastern Europe + 23706.2 24095.8 195.6 207.6 Baltic CIS 24349 24682.3 24578.4 215.8 215.5 208.2 24694 25963.2 26272.4 199 193.3 194.3 26455 25278.4 199.7 517.7 52425 3758.5 48883.8 56347.3 56707.2 59385.4 51802.4 52537.4 52308.5 52503.8 52615.7 1472.4 1452.8 1389.4 1476.2 1722.1 1956.1 2189.7 2474.4 2775.6 Eastern Asia 127360.5 133586 137344.7 141265.6 143299.7 195662.7 210134.8 223499.4 238283.5 492394.3 46480.4 46639.3 46748.5 46764.6 46900.5 47059.3 47338.6 47515.1 47727.8 28416.5 140342.3 144374.8 148917.8 157096.2 162093.4 166006.6 171538.6 177894.5 184730.3 274556.5 6930.3 7200.7 7599.2 8472.6 8973.9 9429.9 9954.3 10540.7 11208.9 17625.6 South Central Asia South-Eastern Asia 82226.8 77148.9 87036.2 100612.7 116269.4 4573.7 4616.6 5110.4 5202.8 5238 128303 133113.9 136482.6 140031.1 178735.8 5233.6 5219.1 5239.7 5231.8 5002.6 Central America 31004 31997.2 32722.8 426.5 447.3 459.6 34303 34893.8 36274.8 42051.5 43612.2 44757.7 56941.7 467 471.4 480.1 491.4 505.1 516.3 622.5 South America 81457.4 84755.2 87235.1 89628.2 98984.6 93443.2 101489.5 104750.2 107995.5 129352.1 1034.8 1065.9 1088.3 1104.3 1095.1 1127.7 1143.2 1160.1 1170.2 1222.3 78722.8 80557.3 85428.2 90020.2 3397.4 3512.1 3623 3702.4 43513.6 46077.5 1066 1097.2 47881 1123.1 88458 90963.6 93405.2 3812.6 3921.5 4022.3 96065 99408.9 4118.2 4229.3 132476 5663.3 Sub-Saharan Africa Middle East 50297 50713.8 52474.6 53763.2 55468.4 57253.1 77716.6 1163.7 1169.9 1192.5 1218.8 1249.4 1280.3 1608.2 Note: The standard deviation is given below employment estimates 39 Table A.19: Total Employment by Economic Sectors: Agriculture (%) Region Europe 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 9.1 0.4 3.8 0.1 21.7 0.3 27.2 0.5 57.0 6.1 53.6 1.3 44.2 2.0 24.6 0.6 15.6 0.7 48.5 1.6 24.5 1.1 8.7 0.4 3.7 0.1 21.5 0.3 23.9 0.5 56.9 6.1 53.8 1.3 48.2 2.0 25.5 0.6 15.3 0.7 48.7 1.6 23.3 1.1 8.7 0.4 3.6 0.1 21.1 0.3 23.9 0.5 58.7 6.0 53.3 1.4 45.6 2.2 23.4 0.6 14.8 0.7 47.1 1.6 23.1 1.1 8.5 0.4 3.5 0.1 20.7 0.3 21.0 0.5 58.8 6.0 50.6 1.4 39.0 2.2 23.1 0.6 15.4 0.7 46.6 1.6 23.0 1.1 7.6 0.4 3.5 0.1 20.8 0.3 24.0 0.6 58.7 5.9 49.5 1.5 34.0 2.1 22.5 0.6 7.4 0.7 48.2 1.6 25.0 1.0 7.1 0.4 3.3 0.1 20.7 0.3 23.3 0.6 48.7 5.9 49.5 1.5 28.6 2.1 21.7 0.6 16.2 0.7 48.1 1.6 25.0 1.0 7.9 0.4 3.0 0.1 14.3 0.3 24.3 0.7 46.7 5.8 48.3 1.5 26.4 2.0 15.9 0.6 8.5 0.7 48.0 1.6 24.9 1.0 7.8 0.4 3.0 0.1 13.6 0.3 23.9 0.7 45.1 5.8 46.9 1.6 26.0 2.0 15.0 0.6 8.0 0.7 47.9 1.6 24.5 1.0 7.5 0.4 2.8 0.1 12.9 0.3 23.5 0.8 43.3 5.7 45.2 1.6 25.5 2.0 14.9 0.6 7.9 0.7 47.6 1.6 24.2 1.0 5.4 0.3 1.9 0.4 10.8 0.8 24.2 1.0 12.8 3.0 31.6 1.6 17.6 1.5 13.2 0.6 6.5 0.5 47.2 1.7 20.0 0.9 Major Non-Europe Eastern Europe + Baltic CIS Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Note: The standard deviation is given below employment estimates 40 Table A.20: Total Employment by Economic Sectors: Industry (%) Region Europe 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 29.1 0.1 26.5 0.6 33.3 0.2 32.3 1.1 26.1 0.7 15.4 0.8 18.7 0.5 22.5 0.3 22.4 0.3 12.8 0.3 25.7 0.5 28.9 0.1 26.5 0.6 33 0.2 29.1 1.1 25.6 0.7 15.3 0.8 17.6 0.5 22.2 0.3 22.4 0.3 12.8 0.3 25.7 0.5 28.8 0.1 26 0.6 32.8 0.2 28.4 1.1 23.4 0.7 15.6 0.8 16.1 0.4 24.6 0.4 22 0.3 13.2 0.3 25.6 0.5 28.3 0.1 25.5 0.8 32 0.2 27.1 1.2 23 0.7 17.3 1 17.9 0.4 23.7 0.4 21 0.3 12.7 0.3 25.7 0.5 28.3 0.1 25.2 1 31.5 0.2 31.8 1.4 23.1 0.7 18.3 1 17.7 0.4 24.5 0.4 23.3 0.3 12.9 0.3 24.4 0.5 27.9 0.1 24.6 0.9 31.5 0.2 32.4 1.5 26.8 0.7 18.2 1.1 19.1 0.4 23.9 0.4 20.5 0.3 12.9 0.3 24.3 0.5 27.8 0.1 23.8 1 35.2 0.2 31.9 1.7 27.3 0.7 19 1.1 20.1 0.4 22.2 0.4 23.7 0.3 13 0.3 24.5 0.5 27.8 0.1 23.6 1.1 35.6 0.3 32.5 1.9 27.6 0.7 19.9 1.2 20.3 0.5 22.2 0.4 23.8 0.3 13 0.3 24.8 0.5 27.7 0.1 23.2 1.2 36 0.3 33 2.2 27.9 0.7 21.1 1.3 20.7 0.5 22.2 0.4 23.8 0.3 13.1 0.3 25.1 0.5 27.1 0.1 16.7 3.2 39.9 0.7 33.3 2.9 32.1 0.9 28.7 2 26.2 0.6 21.8 0.4 24.8 0.4 13.1 0.3 28.2 0.6 Major Non-Europe Eastern Europe + Baltic CIS Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Note: The standard deviation is given below employment estimates 41 Table A.21: Total Employment by Economic Sectors: Services (% Region Europe 1996 1997 1998 1999 2000 2001 2002 2003 2004 2015 61.8 0.4 69.7 0.6 45.0 0.4 40.4 1.2 16.9 6.2 31.0 1.5 37.1 2.1 53.0 0.7 62.1 0.8 38.7 1.7 49.8 1.2 62.4 0.4 69.8 0.6 45.5 0.4 47.0 1.2 17.5 6.1 30.9 1.5 34.2 2.0 52.3 0.7 62.3 0.8 38.5 1.7 51.0 1.2 62.6 0.4 70.4 0.6 46.1 0.4 47.8 1.2 17.9 6.1 31.1 1.6 38.3 2.3 52.0 0.7 63.2 0.8 39.7 1.7 51.3 1.2 63.2 0.5 71.0 0.8 47.2 0.4 51.9 1.3 18.2 6.0 32.1 1.7 43.1 2.2 53.2 0.7 63.6 0.8 40.7 1.7 51.2 1.2 64.2 0.4 71.3 1.0 47.7 0.4 44.2 1.5 18.2 6.0 32.3 1.8 48.3 2.2 53.0 0.7 69.3 0.8 38.9 1.7 50.6 1.2 65.0 0.5 72.1 0.9 47.8 0.4 44.2 1.6 24.6 5.9 32.3 1.8 52.3 2.1 54.4 0.7 63.3 0.8 38.9 1.7 50.7 1.2 64.4 0.4 73.2 1.0 50.5 0.4 43.8 1.8 26.0 5.9 32.7 1.9 53.5 2.1 61.9 0.7 67.8 0.8 39.0 1.7 50.6 1.1 64.5 0.4 73.4 1.1 50.8 0.4 43.6 2.1 27.3 5.8 33.2 2.0 53.7 2.1 62.7 0.7 68.2 0.8 39.1 1.7 50.6 1.1 64.7 0.4 73.9 1.2 51.1 0.4 43.5 2.3 28.8 5.8 33.7 2.0 53.9 2.0 62.9 0.7 68.3 0.7 39.3 1.7 50.7 1.1 67.5 0.3 81.3 3.3 49.4 1.0 42.5 3.0 55.1 3.2 39.7 2.5 56.2 1.6 65.0 0.7 68.7 0.6 39.7 1.7 51.8 1.1 Major Non-Europe Eastern Europe + Baltic CIS Eastern Asia South Central Asia South-Eastern Asia Central America South America Sub-Saharan Africa Middle East Note: The standard deviation is given below employment estimates EMPLOYMENT STRATEGY PAPERS 2004/1 2004/2 Macroeconomic reforms and a labour policy framework for India, by Jayati Ghosh Macroeconomic reforms, labour markets and labour policies: Chile, 1973-2000, by Guillermo Campero Employment and labour market effects of globalization: Selected issues for policy management, by Haroon Bhorat and Paul Lundall Successful employment and labour market policies in Europe and Asia and the Pacific, by Claire Harasty (ed.) Global poverty estimates and the millennium goals: Towards a unified framework, by Massoud Karshenas The labour market effects of US FDI in developing countries, by Robert E. Lipsey Industrial relations, social dialogue and employment in Argentina, Brazil and Mexico, by Adaberto Cardoso Global employment trends for women, 2004, by Sara Elder and Dorothea Schmidt Agricultural productivity growth, employment and poverty in developing countries, 1970-2000, by D.S. Prasada Rao, Timothy J. Coelli and Mohammad Alauddin Efectos de la apertura comercial en el empleo y el mercado laboral de M xico y sus diferencias con Argentina y Brasil (1990-2003), by Enrique Dussel Peters Capital inflows and investment in developing countries, by Ajit K. Ghose Reaching Millennium Goals: How well does agricultural productivity growth reduce poverty?, by Nomaan Majid Labour market policies and regulations in Argentina, Brazil and Mexico: Programmes and impacts, by Adriana Marshall Estimating growth requirements for reducing working poverty: Can the world halve working poverty by 2015?, by Steven Kapsos Insights into the tenure-productivity-employment relationship, by Peter Auer, Janine Berg and Ibrahima Coulibaly Imputation, estimation and prediction using the Key Indicators of the Labour Market (KILM) data set, by Gustavo Crespi Tarantino 2004/3 2004/4 2004/5 2004/6 2004/7 2004/8 2004/9 2004/10 2004/11 2004/12 2004/13 2004/14 2004/15 2004/16 These papers may be downloaded from our website at www.ilo.org/public/english/employment/strat/espapers.htm

Find millions of documents here - Study Guides, Homework Solutions, Papers, Exam Answer Keys and more. Course Hero has millions of course related materials that will enable you to learn better, faster and get an A in all your courses.
Below is a small sample set of documents: