Measuring underemployment: does the cut-off point really matter?

Abstract

Unemployment and underemployment are the most pressing problems in Asia today, as reflected in the widespread underutilization rate of about 29% of the total labor force. Most labor forces in developing countries cannot afford to be completely unemployed and the standard labor force framework currently in use worldwide is biased toward counting the labor force as employed rather than unemployed. This systematically undervalues the full extent of the unemployment problem. The underemployment indicator is then introduced to overcome the issue but the existing guidelines for measuring time related-underemployment using the cut-off point for working full-time, set the threshold too low, resulting in the considerable under-representation of underemployment. This paper suggests a better way of determining the threshold using the cluster method. The robustness of its results is assessed and the overall results suggest that the proposed cut-off point of 40 working hours per week is the best one.

Keywords:

• underemployment
• underutilization
• cut-off point
• full-time work
• measurement
• statistical and econometric assessments JEL Classifications: C12
• C42
• E24
• J08
• J2
• J82

Acknowledgements

The views expressed in this paper are those of the author and do not necessarily reflect the views and policies of the Asian Development Bank, or the Board of Governors or the governments they represent. Eric B. Suan of ERD/ADB provided excellent research assistance for this paper, and the author also thanks the reviewer for valuable comments.

Notes

1. The reference period of a labor force survey is usually one week prior to the survey date.

2. Employers, own-account workers, and unpaid family workers are considered as self-employed.

3. Persons temporarily not at work because of illness or injury; holiday or vacation; strike or lock-out; educational or training leave; maternity or parental leave; reduction in economic activity; temporary disorganization or suspension of work due to such reasons as bad weather, mechanical or electrical breakdown, or shortage of raw materials or fuels; or other temporary absence with or without leave should be considered as in paid employment provided they have a formal job attachment.

4. The economically active population comprises all persons providing the supply of labor for the production and processing of economic goods and services for the market, for barter, or for own consumption.

5. The Hours of Work Convention, 1919, No 1 (Industry)(ILO 2007a); and the Hours of Work Convention, 1930, No. 30 (Commerce and Offices) (ILO 2007b).

6. This is reflected in the Forty-Hour Week Convention, 1935, No 47; and the Reduction of Hours of Work Recommendation, 1962, No. 116.

7. Altogether, the ILO Manual on Key Indicators of the Labor Market consists of seven major dimensions of a labor market, including indicators on labor force, employment, unemployment and underemployment, educational attainment, wages and compensation costs, labor productivity and labor costs, poverty and income distribution (ILO 2007).

8. However, factor analysis has an underlying theoretical model, while cluster analysis is more ad hoc.

9. The k-means algorithm was popularized and refined by Hartigan (1975). The basic operation of that algorithm is that given a fixed number of (desired or hypothesized) k clusters, observations are assigned to those clusters so that the means across clusters (for all variables) are as different from each other as possible (see also Hartigan and Wong 1978).

10. This is analogous to ‘ANOVA in reverse’ in the sense that the significance test in ANOVA evaluates between-group variability against within-group variability when computing the significance test for the hypothesis that the means in the groups are different from each other. In k-means clustering, the program tries to move objects (e.g. cases) in and out of groups (clusters) to get the most significance ANOVA results.

11. For more information about the survey including some of its main results, see the BPS website at http://www.bps.go.id/sector/employ/index.html.

12. This is a statistical term meaning that the data distribution is skewed to the right of the graph, and not to the right-hand side of the reader.

13. This number is slightly different from the number of workers working less than 40 hours per week presented in Figure 6, which is around 46 million workers. The main reason is because the calculation of the number of workers during 1990–2003 in Figure 6 is based on the unrevised version of SAKERNAS 2003, while this study is based on the revised version.