STATISTICAL STUDY OF UNEMPLOYMENT IN SRI LANKA. Bothalegama Hemanthi Pushpa Kumari

Transcription

1 STATISTICAL STUDY OF UNEMPLOYMENT IN SRI LANKA 33 Bothalegama Hemanthi Pushpa Kumari Thesis submitted to the University of Sri Jayewardenepura for the award of the Degree of Master of Science in Applied Statistics in 2006.

2 Declaration The work described in this thesis was carried out by me under the supervision of Mr. P. Dias and a report on this has not been submitted whole or in part to any university or any other institution for another Degree/Diploma. B. H. Pushpa Kumari Reg. No: GS/PS Index No: PGD/PS/0084

3 Declaration I certify that the above statement made by the candidate is true and that this thesis is suitable for submission to the University for the purpose of evaluation..... Mr. P. Dias, (Supervisor) Senior Lecturer, Department of Statistics and Computer Science, University of Sri Jayewardenepura.

4 Table of Contents Page No. Acknowledgements... VIII Abstract... IX-X CHAPTER 1- INTRODUCTION... I 1.1 Labour Force Surveys in Sri Lanka Field Work of the Survey Sample Design of the Survey Probabilities of Selection Literature review Objectives of the study Significance of the study Limitations of the study Concepts and Definitions Methodology Recoding the data Sofiwares used in the study Organization of the report I

5 CHAPTER 2- PRELIMINARY ANALYSIS OF THE LABOUR FORCE SURVEY CHAPTER 3- DATA ANALYSIS Introduction Association of Employment Status with Sector Association of Employment Status with Ethnic Group Association of Employment Status with District Association of Employment Status with Gender Association of Employment Status with Religion Association of Employment Status with Voc\Tech Training Association of Employment Status with whether the Voc\Tech Training Formal! Informal Association of Employment Status with Educational Attainment Association of Employment Status with Age Group

6 3.11Association of Employment Status with Marital Status Conclusion...33 CHAPTER 4- EXPRESSING EMPLOYMENT RATIO IN TERMS OF SOME INFLUENTIAL FACTORS Introduction Methodology Model Assumptions Adequacy of the model Comparison of two models Analysis Conclusion CHAPTE 5- CONCLUSIONS AND DISCUSSION Introduction Conclusions Discussion...46 CHAPTER 6- RECOMMENDATIONS Introduction Recommendations...47 III

7 REFERENCES. 48 APPENDICES. 49 Appendix I Appendix Iv

8 List of Tables Table 2.1- Sample size related to the survey period...14 Table 2.2- Distribution of sample by Sector Table 2.3- Variation of sample with respect to District Table 2.4- Contribution of sample by Gender Table 2.5- Variation of sample by Ethnic Group Table 2.6- Distribution of sample by Religion Table 2.7- Number of persons Employed, Unemployed and Economically Inactive Table Composition of Economically Active by Sector Table Composition of Economically Active by Ethnic Group 21 Table Composition of Economically Active by District Table Composition of Economically Active by Gender Table Composition of Economically Active by Religion Table Composition of Economically Active by Religion and Gender Table Composition of Economically Active by Voc\Tech Training...27 Table Composition of Economically Active by whether the Voc\Tech Training Formal! Informal...28 Table Composition of Economically Active by Educational 29 Attainment... V

9 Table Composition of Economically Active by Age Group...30 Table Table Composition of Economically Active by Marital Status...32 Summary of the output at step 1 in Generalized Linear Models...39 Table Summary of the output at step 2 in Generalized Linear Models...40 Table Significant Variables and factor levels...42 VI

10 List of Figures Chart 2.1- Distribution of sample by Sector. 15 Chart 2.2- Graph 2.1- Distribution of sample by Ethnic Group...17 Unemployed rates by sex-census of population andhousing, Graph Unemployed percentage with respect to district 23 Graph Unemployed Percentage with respect to Educational Attainment...30 VII

11 Acknowledgements I wish to express my sincere gratitude to Mr. P. Dias senior lecturer in the Department of Statistics and Computer Science for supervising this project and turning me on to the correct path. I wish to extend my sincere thanks to Dr. B. M. S. G. Banneheka Coordinator/Postgraduate DiplomalM. Sc. in Applied Statistics who gave me valuable advice throughout the whole project. My special thanks to Mr. D. C. Wickrarnarachchi lecturer of the Department of Statistics and Computer Science for sharing his data analyzing experience with me. I take this opportunity to thank staff members of Department of Census and Statistics who helped me to obtain data and gave other important details which supportive for the project. With gratitude I remember Mr. and Mrs. Paris who gave me fullest cooperation to complete the thesis. My special thanks to my friends Miss. H.M.I Kumari, Miss. K.G.D.C. Wijewardhana, Miss. S. Jayasundara and Mr. J.M.K.H. Bodhiratna who helped me immensely in my study. VIII

12 A STATISTICAL STUDY OF UNEMPLOYMENT IN SRI LANKA Bothalegama Hemanthi Pushpa Kumari ABSTRACT Labour Force Surveys provide estimates of employment and unemployment which are most important measures of performance of a country's economy. The aim of this study was to further analyze the factors relating to the unemployment in Sri Lanka. The report presents findings based on the data extracted from Sri Lanka Labour Force Survey-2003, Conducted by the Department of Census and Statistics. Department has used stratified two-stage probability sampling technique to select the sample. The result of the statistical analysis was that Employment Status of a person is associated with the Gender of the person, Religion, Marital status, Age of the person, Ethnic group, Place of birth (District), Sector (Urban, Rural, Estate), Educational Attainment, Whether the selected person had a Vocational/Technical training and Vocational/Technical training formal or informal. According to the study unemployed rate is higher, for the people who had Vocational/Technical training than persons who had not got Vocational/Technical training. The reason may be persons who had Voc/Tech training are waiting for white collar jobs or mismatch between the skills provided by Sri Lanka's Ix

13 VocationallTechnical training institutions and the skills demanded by the industry and the service sector. This study constructs a model which Express the Employment ratio in terms of some influential factors. x

14 CHAPTER 1 Introduction 1.1 Labour Force Surveys in Sri Lanka Labour force is a main resource of a developing country like Sri Lanka. To make effective future plans, a country should have an in depth knowledge and understanding about its Labour Force and obstacles connected to it. Detailed statistical information on unemployment is very important to a country for several reasons. These statistics clearly indicate the number of persons who are not employed during a specific period. Therefore analysis of unemployment information presents a better insight into the economic health of a country. Labour force surveys are essential to obtain necessary information. The last labour force survey was conducted in 1985/86. The variables measured in those previous surveys were not enough to give valuable information to the country. (Sri Lanka Labour Force Survey Sample Design-Departeinent of Census and Statistics) Therefore the Department of Census and Statistics of Government of Sri Lanka was interested in conducting regular ongoing household survey to obtain up to date and accurate labour force information on quarterly basis. The survey is being conducted quarterly since the first quarter of Nationally representative samples of about housing units are selected annually. From that annual sample about 4000 housing units were enumerated quarterly. This survey was designed to measure the levels and trends of employment and labour force in Sri Lanka. I

15 1.2 Field Work of the Survey The fieldwork of the survey in 2003 was done in quarterly basis. Together the maps and listing records (Sampling frame obtained from the population census -2001) provided a complete account of all housing units within the sample while the maps provided visual identification of the block boundaries and housing units on the ground. The list provided such detailed information as the name of head of household, street address, description and occupancy status Sample Design of the Survey A stratified two stage probability sample design used with census block (CB) as primary sampling units (psu's) and housing units (HH) as secondary and final sampling units. (ssu's) There are three levels of stratification. The 9 provinces were divided into urban and rural sectors and the Greater Colombo area was considered separately. (Greater Colombo includes Colombo Municipal Council, Dehiwala Mount Lavinia Municipal Council and Kotte Municipal Council.) In the rural area, estate and non estate substrata were considered separately. (The estate sector is composed of large plantations of tea, rubber coconut and other crops.) Districts were considered separately in the third phase. 2

16 1.2.2 Probabilities of Selection The sample was divided into four independent sub samples which were assigned at random to the four quarters of the year. Quarterly provided only estimates only on national levels h Substratum within an estimation domain, defined by the province! sector I sub sector/ district cell The selected primary sampling unit (psu) specifically the sample census block; i j The sample household j1,...,m 11, N b Total number of sample blocks within substratum h (From the 2001 census) n h Number of sample blocks selected in substratum h M 11 Total number of housing units in substratum h from the 2001 census M, Total number of housing units in the ith sample block of substratum h from the 2001 census. M, Number of eligible housing units listed in the 1h sample block of substratum h. in, Uniform number of sample housing units in the it1' substratum h. Within each substratum h, the probability of selection of jib psu (block) Pr(block) = Mb The probability of selection of the j th household given the inclusion of the i " block in the sample. 3

17 Pr(HH!block) = M' h, There for the over all probability of selection of the j " household is Pr (block) Pr (HHIblock) = n,m, Il in Mh M'h, The base sampling weight for each household in the sample is given by the reciprocal of its probability of selection. Then the weight W = Mh M hi nh A'i hl mhl Source: Labour Force Survey Sample Design-Department of Census and Statistics) 1.3 Literature review Nanayakkara A.G.W (2004) published a report 'Employment and Unemployment in Sri Lanka - Trends, Issues and Options', which explains the Employment and Unemployment issues. The main emphasis in this report was on the unemployment situation in the country. An attempt was also made to identify specific issues related to employment and unemployment. He also suggested certain policy options which may help in rectifying some of the problems. The main data source for the analysis was provided from the quarterly Labour Force Survey which is being conducted in each quarter by the Department of Census and Statistics since the first quarter of The data which were collected during 1990 to 2002 have been used for the analysis. According to his analysis some of the major issues related to employment and unemployment situations as follows. El

18 Around 80% of the unemployed are in the age range from 15 to 29 and around 60% are in the age range from 20 to 29 years. The unemployment rate for the latter which is the worst affected group had been ranging from around 30% in 1990 to 20% in The unemployment rate is highest among educated youth. The rate has decreased from around 30% in 1990 to 17% in 2002 for those who have GCE (AlL) or above. However, the unemployment rate is still higher for educated youth, than the less educated groups. The situation is worse for females compared to males. The unemployment is not restricted only to educated youth but also to less educated. Around 40 to 50 percent of the unemployed have completed only grade In the case of males, around 50 to 58 percent are in this group. Unemployment rate is highest in the Southern province, throughout the past few zd decades. Sabaragamuwa province also has not shown an improvement where as other province have shown slight improvement. Graduate unemployment has already become a serious problem in this country and may become worse in the future, as the numbers are increasing each year. Dharma Dissanayake (1999) has done a study on 'Changing Pattern of Labour Force She has done this study to compare the changes of the labour force characteristics in 1992 and The two surveys conducted during the 12 months sub rounds in each year and based on the same sample using same questionnaire She compared the unemployed rate in 1992 and 1997 with respect to age, sex, level of education, sector, district and age group. 61

19 Department of census and statistics published Labour Force Survey reports quarterly and annually since Following results can be obtained from the reports. Estimated unemployed population of the country. Duration of unemployed with respect to education level and sex Percentage of unemployed by sex, level of education, district, sector, age group, Rama and Martin (The World Bank Group) released a report under the title 'The Sri Lankan unemployment problem revisited'. According to their study they said Sri Lanka's high unemployment rate has been attributed to a mismatch of skills, to queuing of public sector jobs, and to stringent job security regulations. But the empirical evidence supporting these explanations is weak. The author takes a fresh took at the country's unemployment problem, using individual records from the 1995 Labour Force Survey and time series for wages in the economy's formal and informal sectors. He has found substantial rents associated with jobs in the public sector and in private sector activities protected by high tariffs or covered by job security regulations. From a time series analysis he has found that most of the unemployed are waiting for 'good' job openings but are not interested in readily available 'bad' jobs. Also they said that unemployed problem is not a shortage of jobs but the artificial gap between good and bad jobs.

20 1.4 Objectives of the study Following are the main objectives of the study. Find significant factors associated with Employment Status of a person. (Employment Status whether a person employed or unemployed) Expressing Employment ratio in terms of some influential factors. 1.5 Significance of the study Unemployment, poverty and malnutrition are common problems to many countries all over the world including Sri Lanka. To have a well being social system 'Unemployment rate' should be low. A person to be 'unemployed' means that person able to work but unable to find work. My study concentrates on finding the significant factors relating to a person to be unemployed. All of the above publications (see 1.3) limited to their analysis up to descriptive level. So I thought it is very important to do a further analysis using statistical techniques and find factors relating to 'Unemployment' significantly. 1.6 Limitations of the study Statistical data relating to the year 2004 could not be used, as data was incomplete due to none coding. Therefore the available complete data for the year 2003 was utilized. Language skills have a direct impact on employment status as experienced currently. However this aspect could not be analyzed since the related information was not included in the questionnaire. If a well qualified person had been employed through out but if he had been unemployed one year prior to the survey he too would be considered as an unemployed ':1

21 person. If the survey had been carried out whilst the said person been employed, he would be considered as an employed person. Therefore the same factors could be considered for employed as well as unemployed in the statistical analysis depending on the time of data collection. 1.7 Concepts and Definitions Labour Force- The labour force is composed of the economically active population 10 years of age and over. The Economically Active Population is defined as those persons who are/were employed or unemployed during the reference period of the survey. Employed- Persons, who during the reference period, worked as paid employees, employers, own account workers (self employed), or unpaid family workers are said to be employed. This included persons with a job but not at work during the reference period. Unemployed- Persons available and/or looking for work, and who did no work, during the reference period are said to be unemployed. Reference period- In this survey two reference periods, "current" and "usual". The "current" reference period is the week preceding the week of the survey, i.e. last week. The "usual" reference period is the preceding 12 months, i.e. the past 12 months. Currently Economically Active- A person who was employed or unemployed during the current reference period is considered to be currently economically active.

22 Usually Economically Active- A person whose major activity during the past 12 months was employment and/or unemployment is considered to have been usually economically active. Currently Employed- A person who worked for pay, or unpaid family gain for one hour or more during the past week is said to be currently employed. Usually Employed- A person whose major economic activity during past 12 months was employment is said to have been usually employed. Currently Unemployed- A person who did no work but was looking for work during the past four weeks is said to be currently unemployed. Usually Unemployed- A person who was looking or available for work during the major part of the past 12 months is said to have been usually unemployed. Not in the Labour Force (not economically active)- Persons who were neither working not available/looking for work are classified as "not in labour force". Persons are not in the labour force for such reasons as: full time care of the household, full-time students, retired or old age, infirmed or disabled, or are not interested in working for one reason or another. 1.8 Methodology Secondary data which were gathered by Department of Census and Statistics for the Labour force survey 2003 was used for this analysis. In their survey the Department used a questionnaire to collect data. There were forty two questions in that questionnaire. But according to my objectives I have selected eleven questions from the questionnaire. OJ

23 The variables selected are as follows. Whether the selected person employed or unemployed during the past twelve months (With respect to the date of survey.) Gender of the person Religion Marital status Age of the person Ethnic group Place of birth (District) Sector (Urban, Rural, Estate) Educational Attainment Whether the selected person had a Voc/Tech Training II. Voc/Tech training formal or informal Recoding the data According to the definitions of Department of Census and Statistics the labour force is defined as composed of economically active population (those persons who are/were employed or unemployed during the reference period of survey.) 10 years of age and over. To facilitate the analysis the selected sample was arranged in 10 year age groups starting from year 10. Marital status was classified into five groups. But observed frequencies in divorced and separated were not large enough to perform a chi-square test. So I combined them in to one group. 10

24 The questionnaire issued by Census and Statistics, the educational attainment of a person is presented in seventeen categories. But in my analysis these seventeen categories were grouped into ten categories as follows. Educational Attainment Recoded Studding Year I Passed Year 1 Passed Year 2/Grade I Jp to year 5 Passed Year 3/Grade 2 Passed Year 4/Grade 3 Passed Year 5/Grade 4 Passed Year 6/Grade 5 Passed Year 7/Grade 6 ear 6-8 Passed Year 8/Grade 7 Passed Year 9/Grade 8 Year 9-10 Passed Year 10/Grade 9 Passed year I 1/G.C.E.(U.L)/N.C.G.E Year 1 1/G.C.E.(O.L)/N.C.G.E Passed Year 12/Grade 11 Up to year 12 Passed Year 13/ G.C.E(A.L)/H.N,C.E Passed GAQ/GSQ Degree Post Graduate Degree/Diploma No Schooling G.C.E(A.L)/H.N.C.E GAQ/GSQ Degree Post Graduate Degree/Diploma No Schooling

25 1.8.2 Softwares used in the study I used following softwares for the study.. MINITAB 12.0 SPSS Organization of the report A factual findings and the conclusion drawn from the study is presented under 6 chapters. The objectives, areas covered scope of work and the conclusion reached under each chapter is depicted as mentioned below. Chapter 1 - This chapter is intended to explain the selection of sample, literature review, objectives, limitations, significance of the study and methodology. Chapter 2 - This chapter can be considered as the first screening analysis of the report which does not give any conclusions but instead provides the basic description of the data set. Chapter 3 This chapter concludes the association of variables which affects Employment Status of a person. Chapter 4 - Under this chapter a model is created to predict the unemployed ratio for which Logistic Regression method was applied. 12

26 Chapter 5 - An over all conclusions is given under this chapter for all the findings which are made in the above mentioned four chapters. Also this chapter gives a brief discussion about the conclusions drawn. Chapter 6 Recommendations 13

27 CHAPTER 2 Preliminary Analysis of the Labour Force Survey Introduction Sri Lanka Labour Force Survey was designed to measure the levels and trends of employment, unemployment and labour force in Sri Lanka. (Quarterly Report of the Sri Lanka Labour Force Survey-Third Quarter-2003-Deparfment of Census and Statistics) Table 2.1-Sample size related to the survey period Month Sample Size Contribution October November ,2 Total According to the Table 2.1, the field work of the survey was done in October and November The survey covers only the non-institutional household population aged 10 years or over and it excludes the institutional population. A household consists of all persons, related or not, who occupy a housing unit, which covers a house, an apartment, a group of rooms when occupied as separate living quarters. Domestic servants, Boarders, and lodgers who live and have their meals or share other essentials of living with the family are also related as members of the household. Persons, who usually live here but are temporary away are includes as member of the household, but persons who have usual residence elsewhere, but are present at the time of the interview, are excluded. (Quarterly Report of the Sri Lanka Labour Force Survey-Third Quarter Department of Census and Statistics) 14

28 Table 2.2-Distribution of sample by Sector Sector Composition Percentage Estate Urban Rural Total Chart 2.1- Distribution of sample by Sector Estate The nine provinces were divided into Urban, Rural and Estate sectors. Table 2.2 and Chart 2.1 represents sample representation withrespect to sectors. The highest participation rate (74%) has been reported from rural sector, while the lowest participation rate (7%) is represented from Estate sector. 15

29 Table 2.3-Variation of sample with respect to District District Composition Percentage I Colombo Gampaha Kurimagala Kandy Kalutara NEliya Kegalle Anuradhapura Ratnapura Baddula Matara Galle Hambanthota Ampara Matale Puttlam Monaragala Polonnaruwa Batticaloa Trincomalee Total The highest participation is reported from Colombo district, while the lowest participation is reported from Tricomalee district. This survey was not covered only the Northern Province. Table 2.4-Contribution of sample by Gender Gender Number Percentage Male Female Total This table depicts the male and female representative rates of the sample. There is no marked difference between male and female rates of the sample. 16

30 Table 2.5-Variation of sample by Ethnic Group Ethnic Group Frequency Percent % of census at 2001 Sinhalese Sri Lanka Moor Indian Tamil Sri Lanka Tamil Malay Burgher Other Total Chart 2.2- Distribution of sample by Ethnic Group Table 2.5 shows that Sri Lankan population is well represented by the sample. Table 2.6-Distribution of sample by Religion Religion Frequency Percent % of census at 2001 Buddhist Hindu Muslim Roman Catholic/Other Christian Other Total iviost or me mna1ese are Buddhists. Also the highest contribution is Buddhist, which is 73.0% of the sample. 17

31 Table 2.7- Number of persons Employed, Unemployed and Economically Inactive Composition Number Percentage Usually employed Usually unemployed Economically inactive Total Only 48.1% of the sample have used for the analysis. (Economically active persons) Table Composition of Economically Active persons in the sample Composition of Economically Active Number Percentage Usually employed Usually unemployed Total It can be seen that 83% of the economically active populations are unemployed. Graph 2.1- Unemployed rates by sex-census of population and housing, io A A A A Unemploent rate Both Sexes -A- - A A A- A Unemployment ' 4 *.. 4- rate Male A A A - Unemployment rate Female 0 District 18

32 Graph 2.1 represents the unemployment rates by sex. In each districts female unemployment rate is higher than males. Highest and lowest unemployment rates can be seen in Matara and Colombo districts respectively. But when we considering unemployment rates sex wise it changed with the districts. For males lowest unemployment rate in Monaragala district for females that rate in Nuwara Eliya district. Also highest unemployment rate for males and females can be seen in Galle and Ampara districts respectively.

33 CHAPTER 3 Data Ana'ysis 3.1 Introduction In this chapter association between Employment Status and the interested factors were tested using Chi-Square test. Individual factors were considered separately. 3.2 Association of Employment Status with Sector Table Composition of Economically Active by Sector Composition of Sector Economically Active Urban Rural Estate Total Usuallyemployed 1092 (89.95%) 4479 (91.58%) 570 (95.96%) 6141 Usually unemployed 122 (10.05%) 412 (8.42%) 24 (4.04%) 558 Total 1214 (100.00%) (100.00%) 594 (100.00%) Chi-Sauare Tests Value df Asymp. Sig. (2-sided) Pearson Cu-Square 19.07r Likelihood Ratio Linear-by-Linear Association N of Valid Cases I-Is: Sector of a person does not depend on Employment Status. Vs H1: Sector of a person depends on Employment Status. 20

34 According to Pearson Chi-Square value given above we can reject H0 at 5% level of significance. (P-Value = 0.000). We can conclude the sector where the person lives is related with the Employment Status of a person. Further by looking at unemployed percentage values (Table 3.2.1) unemployed rate is less in estate sector and it is very high in urban sector. 3.3 Association of Employment Status with Ethnic Group Table Composition of Economically Active by Ethnic Group Composition of Economically Active Ethnic Group Usually employed Usually unemployed Total Sinhalese 4814 (91.59%) 442 (8.41%) 5256 Sri Lanka Tarnil 344 (89.12%) 42 (10.88%) 386 Indian Tamil 532 (96.73%) 18 (3.27%) 550 Sri Lanka Moor 432 (89.26%) 52 (10.74%) 484 Malay/Burgher/Other 19 (82.61%) 4 (17.39%) 23 Total 6141 (91.67%) 558 (8.33%) 6699 Chi-Square Tests Value df Asymp. Sig. (2-sided) Pearson Chi-Square a Likelihood Ratio Linear-by-Linear Association N of Valid Cases a. 1 cells (10.0%) have expected count less than 5. The minimum expected count is (Siegel- Condition Il-After combining categories, fewer than 20% of the cells have expecredfrequencies of less than 5 and no cells has an expectedfrequency of less than I meaningfully interpret the result of chi-square test) 21

35 I-1: Ethnicity of a person has no relation with Employment Status. Vs H1: Ethnicity of a person has a relation with Employment Status. According to the Pearson Chi-Square value given above we can reject H. at 5% level of significance (P-Value =0.000). We can conclude that unemployed rates vary with the ethnicity. According to the table given above Indian Tamils have the lowest unemployed rate when comparing with other ethnic groups. The reason may be Indian Tamils have sufficient employments in the state sector. 3.4 Association of Employment Status with District Table Composition of Economically Active by District District Usually employed Usually unemployed Total N'Eliya 467 (97.70%) 11 (2.30%) 478 Kurunagala 508 (95.13%) 26 (4.87%) 534 Kandy 351 (94.86%) 19 (5.14%) 370 Baddula 408 (94.44%) 24 (5.56%) 432 Monaragala 221 (93.64%) 15 (6.36%) 236 Matale 248 (92.54%) 20 (7.46%) 268 Puttlam 208 (92.44%) 17 (7.56%) 225 Gampaha 422 (92.14%) 36 (7.86%) 458 Kalutara 375 (92.14%) 32 (7.86%) 407 Anuradhapura 363 (91.67%) 33 (8.33%) 396 Colombo 513 (91.28%) 49 (8.72%) 562 Galle 299 (91.16%) 29 (8.84%) 328 Ratnapura 327 (90.83%) 33 (9.17%) 360 Trincomalee 53 (89.83%) 6 (10.17%) 59 Kegalle 310 (89.60%) 36 (10.40%) 346 Polonnaruwa 199 (89.24%) 24 (10.76%) 223 Batticaloa 125 (86.81%) 19 (13.19%) 144 Ampara 199 (86.15%) 32 (13.85%) 231 Matara 285 (85.84%) 47 (14.16%) Jambanthota 260 (83.87%) 50 (16.13%) 310 Total 6141 (91.67%) 558 (8.33%)

36 Graph Unemployed percentage with respect to district Unemployed percentage with respect to District , 10 0 H H H H District Chi-Square Tests Value df Asymp. Sig. (2-sided) Pearson Chi-Square Likelihood Ratio Linear-by-Linear A ssociation N of Valid Cases a. 1 cells (2.5%) have expected count less than 5. The minimum expected count is H0: Employment Status has no association with the district where the person resides. Vs H1: Employment Status has an association with the district where the person resides. According to the Pearson Chi-Square test value, we can reject H0 at 5% level of significance. (P-value = 0.000). Therefore, we can conclude that Employment Status has an association with the district of a person. Table gives the unemployed and 23

37 employed percentages out of economically active for each district. Here also a clear variation can be identified. We know that in several times violence started from southern area. Most of educated people interpret that as a result of unemployed younger generation. They were waiting for a job during long period of time. This burning problem changed their minds and ideas. Their minds start to get angry with the whole society and the government. They believe that they can achieve their targets from the violence. Also from the selected sample highest unemployed percentage can be seen from the Matara and 1-lambantota districts. This sample is also in line with above interpretation. 3.5 Association of Employment Status with Gender Table Composition of Economically Active by Gender Gender Composition of Economically Active Male Female Total Usually employed 4228 (93.56%) 1913 (87.75%) 6141 Usually unemployed 291 (6.44%) 267 (12.25%) 558 Total 4519 (100%) 2180 (100%) 6699 Chi-Square Tests Value df Asymp. Sig. (2-sided) Pearson Chi-Square 64972b Continuity Correction a Likelihood Ratio Exact Sig. (2-sided) Exact Sig. (1-sided) Fisher's Exact Test Linear-by-Linear Association N of Valid Cases

38 Ho; Employment Status does not depend on Gender of a person. Vs H1 : Employment Status depends on Gender of a person. According to the percentage unemployed values females are unemployed twice as males. There is a wide difference among gender. Also according to the chi-square value we have enough evidence to reject H0 at 5% level of significance. So a person's gender has an effect on to be employed or unemployed. 3.6 Association of Employment Status with Religion 25 Table Composition of Economically Active by Religion Usually Usually Religion employed unemployed Total Buddhist 4580 (91.404) 431 (8.60%) 5011 Hindu 747 (94.56%) 43 (5.44%) 790 Muslim 465 (89.08%) 57 (10.92%) 522 Roman Catholic/Other Christian 349 (92.82%) 27 (7.18%) 376 Total 6141 (91.67%) 558 (8.33%) 6699 Table Composition of Economically Active by Religion and Gender Relicion Roman Gender Employment Catholic! Other Status Buddhist Hindu Muslim Christian Total Usually Employed % % % % 4208 Usually Male Unemployed % % % % 291 Total % % % % 4499 Employed Usually % % % % 1913 Usually Unemployed % % % % 267 Female Total ,00% % % % 2180 Total

39 Chi-Square Tests Value dl Asymp. Sig. (2-sided) Pearson Chi-Square a Likelihood Ratio Linear-by-Linear Association N of Valid Cases Ho: Employment Status does not depend on Religion of a person. Vs H1 : Employment Status depends on Religion of a person. According to the Pearson Chi-Square value given above we have enough evidence to reject H0 at 5% level of significance. (P-value = 0.002) So we can conclude that Religion has some effect on Employment Status of a person. According to this sample we can find out highest unemployed rate is for Muslim females. That may be due to many of Muslim females are not working as a tradition of their culture. Lowest unemployed rate is seen among Hindu males. Many of the Hindu males work as laborers of plantations. Except for Roman Catholic/Other Christians unemployed rate is varies with sex for all the other religions. Unemployed rate of females is higher for Buddhists, Hindus, and Muslims. But it is very much similar for Roman Catholic/Other Christians. This may be due to cultural barrier in other religions. 26

40 3.7 Association of Employment Status with Voc\Tech Training Table Composition of Economically Active by Voc\Tech Training Usually Usually Voc\Tech Training employed unemployed Total Had a Voc\Tech Training 870 (85.63%) 146 (14.37%) 1016 Had not a Voc\Tech Training 5271 (92.75%) 412 (7.25%) 5683 Total 6141 (91.67%) 558 (8.33%) 6699 Pearson Chi-Square Value 57229b Chi-Square Tests df Asymp. Sig. (2-sided) Continuity Correction a Likelihood Ratio Exact Sig. (2-sided) Exact Sig. (1-sided) Fisher's Exact Test Linear-by-Linear Association N of Valid Cases 6699 a. Computed only for a 2x2 table H0; Employment Status has not an association with whether a person had a Voc\Tech Training or not. Vs H1 : Employment Status has an association with whether a person had a Voc\Tech Training or not. According to the Pearson's Chi-Square value, we have enough evidence to reject at 5% level of significance. (P-Value = 0.000) So whether a person had a voc\tech Training has an association with Employment Status. According to the above table unemployed rate is higher, persons who had Voc/Tech training than persons who had not got 27

41 Voc/Tech training. The reason may be persons who had Voc/Tech training are waiting for white collar job. There is a strong mismatch between the skills provided by Sri Lanka's vocational training institutions and the skills demanded by industry and the service sector. It is estimated that up to 70 percent of vocational courses equip trainees with low-level skills that have little relevance to employment markets. The disparity between the skills taught in vocation institutions and the skills required in the labor market has serious repercussions for Sri Lanka's unemployment problem. (Sri Lanka- ADB.org, 2006) 3.8 Association of Employment Status with whether the Voc\Tech Training Formal! Informal Table Composition of Economically Active by whether the Voc\Tech Training FormaL/Informal Voc\Tech Training Usually employed Usually unemployed Total Formal 654 (83.74%) 127 (16.26%) 781 Informal 216 (91.91%) 19 (8.09%) 235 Total 870 (85.63%) 146 (14.37%) 1016 Chi-Square Tests Asymp. Sig. Exact Sig. Exact Sig. Value df (2-sided) (2-sided) (1-sided) Pearson Chi-Square 9814b Continuity Correctiona Likelihood Ratio Fishers Exact Test.00i.001 Linear-by-Linear Association N of Valid Cases 1016 a. Computed only for a 2x2 table 28

42 Ho: Employment Status does not depend on whether the Voc\Tech Training Formal or Informal. Vs H1: Employment Status depends on whether the Voc\Tech Training Formal or Informal. Here persons who had a formal Voc/Tech training have the highest unemployed rate. This may be due to they are waiting for employment which is relating to their training. Government and other related institutes are not recruiting qualified persons for relevant employments. Unemployed rate is law in persons who had informal Voc\Tech training; the reason may be they tend to go for any job without any selection. 3.9 Association of Employment Status with Educational Attainment Table Composition of Economically Active by Educational Attainment Educational Attainment Usually employed Usually unemployed Total No Schooling 255 (97.70%) 6 (2.30%) 261 Up to Year (97.99%) 27 (2.01%) 1343 Year (95.97%) 49 (4.03%) 1216 Year (91.57%) 151 (8.43%) 1791 Year 1 l/gce(o/l)/ncge 737 (86.10%) 119 (13.90%) 856 Upto Year l2 210 (86.42%) 33 (13.58%) 243 GCE(AIL)II-[NCE 629 (80.23%) 155 (19.77%) 784 GAQ/GSQ 15 (83.33%) 3 (16.67%) 18 Degree 142 (91.03%) 14 (8.97%) 156 Post Graduate Degree/Diploma 30 (96.77%) 1 (3.23%) 31 Total 6141 (91.67%) 558 (8.33%)

43 Graph Unemployed Percentage with respect to Educational Attainment Unemployed Percentage withrespect to Educational Attainment ri U, I o < &., -, A Level of Education Chi-Square Tests Asymp. Sig. Value df (2-sided) Pearson Chi-Square Likelihood Ratio Linear-by-Linear Association N of Valid Cases a. 2 cells (10.0%) have expected count less than 5. The minimum expected count is Ho: Employment Status does not depend on Educational Attainment of a person. Vs H1: Employment Status depends on Educational Attainment of a person. According to the educational attainment we can see lowest unemployed rates distributed among lowest and highest educated people. In this lowest education range represents non schoolings, they prefer to formal or informal job. In middle educated range has 30

44 highest unemployed rate. Because of majority of persons in the country get middle level education. But not generating enough job opportunities to full fill that requirement Association of Employment Status with Age Group Table Composition of Economically Active by Age Group Age Group Usually employed Usually unemployed Total (66.84%) 194 (33.16%) (83.73%) 289 (16.27%) (97.11%) 49 (2.89%) (98.79%) 17 (1.21%) (99.34%) 6 (0.66%) (99.06%) 3 (0.94%) 318 Total (91.67%) 558 (8.33%) 6699 Chi-Square Tests Value df Asymp. Sig. (2-sided) Pearson Chi-Square Likelihood Ratio Linear-by-Linear Association N of Valid Cases H0: Employment Status does not depend on Age of a person. Vs H1: Employment Status depends on Age of a person. We have enough evidence to reject H0 at 5% level of significance. (P-value=0.000) So we can conclude age has an effect on a economically active person to be employed or unemployed. Also we can see unemployed rate varies widely with the age group. The highest percentage of unemployed persons is reported in the age group

45 Legally it is prohibited to be a person to be employed under 14.Normally lowest recruited age greater than Association of Employment Status with Marital Status Table Composition of Economically Active by Marital Status Marital Status Employment Never Married Married Widowed Dov/Sep Total Status No. % No. % No. % No. % Usually employed Usually unemployed Total Chi-Square Tests Value df Asymp. Sig. (2-sided) Pearson Chi-Square a Likelihood Ratio Linear-by-Linear Association N of Valid Cases H0: Employment Status does not depend on marital status of a person. Vs H1 : Employment Status depends on marital status of a person. 32

46 According to the Pearson Chi-Square value we can reject H0 at 5% level of significance. (P-value= 0.000). There for the person to be employed or unemployed depends on marital status Conclusion Gender of the person, Religion, Marital status, Age of the person, Ethnic group, Place of birth (District), Sector (Urban, Rural, Estate), Educational Attainment, Whether the selected person had a Voc/Tech Training and Voc/Tech training formal or informal variables are (i.e. all the interested variables) associated with the Employment Status of a person at 5% level of significance. 33

47 CHAPTER 4 Expressing Employment Ratio in terms of some influential factors 4.1 Introduction Number of employed persons for each covariate class was taken as the response variable and all the other variables were taken as the explanatory variables. The explanatory variables considered are Gender of the person Religion Marital status Age of the person Ethnic group Place of birth.(district) Sector (Urban, Rural, Estate) Educational Attainment Whether the selected person had a Voc/Tech Training Voc/Tech training formal or informal 34

48 4.2 MethodolOgy The random variable Y has a Binomial distribution with parameters n, 8 then Y - I3in(n, 9) The probability density function of y is given by ; yo,i,...,n E(y)=nO and Var(y)n9(19) Canonical link for Binomial distribution is logit link. Therefore, the model would be lne(y) = I XI/3, Where Y is number of employed persons. X 's are explanatory variables. /3 's are coefficients of X1 's Model Assumptions i. Response variable follows a binomial distribution 2. ObservationS are independent. 35

49 4.2.2 Adequacy of the model To check the adequacy of the model, it will be compared with a model that provides a complete description of data. (ie the maximal or saturated model) The saturated model has following properties. It is a generalized linear model with the same distribution and link ftinction as the model of interest. The number of parameters in the maximal model is equal to the number of observations. Suppose that the model of interest is g(1u)=x/3 Where/I is a p-dimensional parameter vector. The maximal or the saturated model g(li) = Z/Imax - Maximum likelihood estimators of the vector of the model of interest. -Maximum likelihood estimators of the vector of regression parameter of the max maximal model. If the model of interest describes the data well L(/3) should be approximately equal to L(!max ) 36

50 U So we can use the likelihood ratio as a measure of goodness of fit. A = max L(/3) Equivalently, the log-likelihood ratio In(A) = l(/3 max ) - 1(f3) could be used. 21n(A) = D (D* is called the scaled deviance) If the model of interest is as nearly adequate as the maximum model and n is large then D Comparison of two models Compare two generalized linear models fitted with same link function. Suppose we want to compare two competing models A and B with respective coefficient vectors. /3 / and /3B= I q<p<n 1; tip Testing hypothesis H0 :/3=,8 Vs Hl:/3=/3R Considering the difference of the log-likelihood ratio AD = - = 21(11 ) ic8)] 2[1(B ) 1(ft )J 37

51 = 2[10 B)_10 A )j If model A is adequate, under H 0, D A %q If model A is adequate then model B also adequate. (Model B contains all the variables in model A and more variables) Therefore, Under H 0, Da Then AD 4.3 Analysis By using Generalized Linear model each explanatory variable was fitted separately. To find the most significant variable G-value was used. Forward selection method was used to find the best fitted model. G value - Difference in two times of log likelihood between a model that only has constant term and the fitted model. Each variable fitted separately to the model. G values and P values for each explanatory variable as follows. 38

52 Table Summary of the output at step 1 in Generalized Linear Models Variable G Df P-value Xl- Sector Sex District Religion HadlVoc Tech Voc/fo/In Edu.Att ,000 Age group X9- Marital Status Xl0- Ethnic Group Highest G value can be seen that model with Age Group (X8) as the explanatory variable. Then we have to find most significant variable when X8 already in the model. Table gives the G values for the models when X8 already in the model.

53 Table Summary of the output at step 2 in Generalized Linear Models Variable G Df P-value Xl X X X X X X X X Highest G value can be seen in the model with X7 and X8. In( + Model (1) 1 ir ' J = In(, = /3 + /31 X 8 + /32 X Model (2) Check whether adding X7 to Model-( 1) significant. Testing hypothesis H 0 : /32 = 0 Vs H1 /2 ~ 0 Difference of G values in Model-(1) and Model-(2) = = 198,414 Change in Degrees of freedom (Df) with respect to above two models = 14-5 =9 Chi square value, % 5% = %95% <

54 So we have enough evidence to reject H0 at 5% level of significance. Then the fitted model is, in' I = + x8 +,2 X7 Above similar procedure was used to find the best fitted model. Best fitted model LVlOf 09] Maritall = + /3h [Age] + fl23 Education + fl3k [District] + / 4m + / 35 [Sex] + fl6 [Voc - tra] ln[1 - I status J = Check the adequacy of the best fitted model. Testing hypothesis H0 : Model is adequate. Vs H1 : Model is not adequate. Goodness-of-Fit Tests Method Chi -Square DF P Pearson According to above p-value there is no evidence to reject H0 at 5% level of significance. Therefore above model is adequate at 5% level of significance. 41

55 Table Significant Variables and factor levels Variable Factor Level Coef Stdev P-Value Age Educational Attainment Year I 1/GCE(OfL) INCGE Up to Year GCE(A/L)/HNCE Degree District Nuwara Eliya Matara Hambantota , Baticaloa Ampara Kegalle Marital Status Married Sex Female Voc/Tech Training Had a Voc/Tech Training (See Appendix I and Appendix 2) 4.4 Conclusion From the interested explanatory variables Age, Level of Education, District, Marital Status, Gender and vocational Training significantly associated with the unemployed ratio at 5% level of significance. When we consider Age Group, unemployed ratio significantly different in all age groups while other factors remaining constant. Out of twenty districts include in the sample significant different unemployed ratio can be seen in Nuwara Eliya, Matara, Hambantota, Baticaloa, Ampara and Kegalle districts. This ratio is significantly different in married persons when comparing with others. Unemployed ratio is significantly different in persons who had Voc/Tech training with persons who had not Voc/tech training at 5% level of significance. 42

56 Difference of unemployed percentage with respect to reference level is as follows. Difference of unemployed % with Variable Factor Level respect to reference level Higher % Lower % Age Reference Level Educational Attainment Year 1 IIGCE(O/L) /NCGE Upto Year l GCE(AJL)/HNCE Degree Reference Level Up to Year 5 District Nuwara Eliya Matara Hambantota Batica!oa Ampara Kegalle Reference District Colombo Marital Status Married Reference Level Never Married Sex Female Reference Level Male Voc/Tech Training Had not Voc/Tech Training Reference Level Had Voc/Tech Training

57 CHAPTER 5 Conclusions and Discussion 5.1 Introduction In this chapter, all the conclusions drawn in the study are reported. Section 5.2 gives the conclusions and section 5.3 gives a brief discussion about the conclusions drawn. 5.2 Conclusions The variables; Gender of the person, Religion, Marital status, Age, Ethnic group, Place of birth (District), Sector (Urban, Rural, Estate), Educational Attainment, whether the selected person had a Voc/Tech Training or whether the Voc/Tech training formal or informal, are(i.e. all the interested variables) associated with the Employment Status of a person at 5% level of significance. Generalized linear model technique was used to model the unemployed rate, and it was found that only Age, Level of Education, District, Marital Status, Gender and Vocational Training are sufficient to explain the unemployed ratio at 5% level of significance. All the age groups have significantly different unemployed ratios. Age group has the highest unemployed ratio when comparing with Age group After age 30 unemployed ratio decreases. However, Age group shows a slight higher unemployed rate when compared with age group