In Pakistan, most of the nationally representative household surveys do not contain information on variables, such as, completed years of schooling, age starting school, literacy and numeracy skills, quality of schooling, and technical training. Due to the unavailability of completed school years, one can neither compute the potential experience nor observe the effect of an additional year of schooling on individual earnings. Therefore, the available literature in Pakistan is lacking in estimating the returns to education by using the Mincerian earning function2.
In recent years, the government of Pakistan has started nation-wide survey, Pakistan Integrated Household Survey (PIHS), to address the imbalances in the social sector. This survey ? The authors are Senior Research Economist and Research Economist at the Pakistan Institute of Development Economics (PIDE) Islamabad. 1 Pascharapoulos (1994) provide a comprehensive update of the estimated rates of returns to education at a global scale. He observed high social and private profitability of primary education (18%and 9% respectively) in all regions of world.
The private rate of returns at this level were found highest in Asia (39%) as compared to other regions. He also noted a considerable increase in total earnings by an additional year of education in all regions of world; 13% in Sub-Saharan Africa; 10% in Asia; 12% in Europe/Middle East/North Africa; and 12% in Latin America/Caribbean. 2 At national level, only two studies are available in Pakistan that used the Mincerian earning function approach to examine the returns to education [see Shabbir and Khan (1991) and Shabbir (1994)].
However, both these studies are based on twenty years old data set. 2 provides rich information on the above mentioned variables that were missing in the earlier household surveys. This study uses the data of PIHS to examine the returns to education by using Mincerian earning function and thus aims to fill the vacuum that, due to the lack of appropriate data, exists in the literature on returns to education in Pakistan. In this paper we will first estimate the earning function with continuous school years with the assumption of uniform rate of returns for all school years.
It is argued that different school years impart different skills therefore we extend our analysis to examine the addition in earning associated with extra years of schooling at different levels of education, i. e. , how much increase in earnings takes place with an extra year of schooling at different levels, such as, primary, middle, matric, intermediate, bachelors and masters. By doing so we overcome the problem that exists in the available literature in Pakistan.
To our knowledge no study has yet adopted this method to examine the returns to education in Pakistan3. The impact of technical training and school quality on the earnings of fixed salaried and wage earners will be examined in this study. Based on the available data in Pakistan, most of the studies, for example, Haque (1977), Hamdani (1977), Guisinger et al (1984), Khan and Irfan (1985), Ahmad, et al (1991); and Ashraf and Ashraf (1993a, 1993b, and 1996) estimated the earning functions by defining the dummy variables for different levels of education4.
These studies observe low rates of returns at different levels of education as compared to other developing countries. However, a positive association between levels of education and earnings and an inverse relationship between the degree of income inequality and educational attainment has been noted. In order to examine the inter- 3 Most of the studies on returns to education in Pakistan used dummy variables for different levels of education where the rates of returns at different levels of education are computed by the estimated coefficients.
4 In Pakistan, the data on education in most of the nationally representative household surveys have been reported in discrete form that denotes the completion of different levels of education, such as, primary but incomplete middle, middle and incomplete matric, and so on. 3 provincial differentials in returns to education, Shabbir and Khan (1991) estimated the Mincerian earning function by using a nationally representative sample, drawn from the of Population, Labour Force and Migration Survey (1979) for the literate wage earners and salaried males.
Later Shabbir (1994) estimated the earning function on the extended sample of the same data set. These studies found 7 to 8 percent increase in earnings with an additional year of schooling. Although the results are consistent with those of comparable LDCs but may not reflect the recent developments in Pakistans economy as these studies are based on the data set which are 20 years old now. Since 1979, the economy of Pakistan has passed through various changes, especially after the inception of the Structural Adjustment Programme in late 1980s.
For example, the literacy rate has increased from 26 percent to 45 percent and enrolment at primary level has increased by 67 percent. Public and household expenditures on education have also increased [Economic Survey (1998-99)]. Moreover, due to the fiscal constraints, the employment opportunities in the public sector have started shrinking and the economy is moving towards more openness with stronger role of private sector in recent years. In this scenario, it becomes imperative to re-test the role of human capital as both private and public sectors are moving towards more efficiency and productivity.
This study is important from three standpoints. First, in order to estimate the effect of education on earnings, the most recent and nationally representative household survey data is used which provides detailed information on the variables that were missing in previous surveys. Second, it uses the splines of education in the earning function to examine the additional earnings associated with extra school years at different levels. Third, this study investigates the role of some important factors such as, technical training, school quality, and literacy and numeracy skills on earnings for the first time.
4 The rest of the paper is organised as follows: section 2 presents an overview of the education sector. Section 3 outlines the model for empirical estimation and describes data. Section 4 reports the results. Conclusions and policy Implications are presented in the last Section. II. The Education Sector in Pakistan: An Overview: Education plays an important role in human capital formation. It raises the productivity and efficiency of individuals and thus produces skilled manpower that is capable of leading the economy towards the path of sustainable economic development.
Like many other developing countries, the situation of the education sector in Pakistan is not very encouraging. The low enrolment rates at the primary level, wide disparities between regions and gender, lack of trained teachers, deficiency of proper teaching materials and poor physical infrastructure of schools indicate the poor performance of this sector. The overall literacy rate for 1997-98 was estimated at 40 percent; 51 percent for males and 28 percent for females; 60 percent in urban areas and 30 percent in rural areas. These rates are still among the lowest in the world.
Due to various measures in recent years, the enrolment rates have increased considerably. However, the high drop-out rate could not be controlled at primary level. Moreover, under-utilisation of the existing educational infrastructure can be seen through low student-institution ratio, (almost 18 students per class per institution) low teacher-institution ratio (2 teachers per institution) and high studentteacher ratio (46 students per teacher). The extremely low levels of public investment are the major cause of the poor performance of Pakistans education sector.
Public expenditure on education remained less than 2 percent of GNP before 1984-85. In recent years it has increased to 2. 2 percent. In addition, the allocation of government funds is skewed towards higher education so that the benefits of public subsidy on education are largely reaped by the upper income class. Many of the highly educated 5 go abroad either for higher education or in search of better job opportunities. Most of them do not return and cause a large public loss. After mid-1980s, each government announced special programs for the improvement of the education sector.
However, due to the political instability, none of these programs could achieve their targets. The Social Action Program was launched in early 1990s to address the imbalances in the social sector. This program aims to enhance education; to improve school environment by providing trained teachers, teaching aids and quality text books; and to reduce gender and regional disparities. The Phase-I of SAP (1993-96) has been completed and Phase-II is in progress. The gains from the Phase-I are still debatable because the rise in enrolment ratio has not been confirmed by the independent sources.
Irrespective of this outcome, government has started work on Phase-II of SAP. In this Phase, government is paying special attention to promote technical and vocational education, expanding higher education in public as well as in the private sector, enhancing computer literacy, promoting scientific education, and improving curriculum for schools and teachers training institutions in addition to promoting primary and secondary education. Due to low levels of educational attainment and lack of technical and vocational education, Pakistans labour market is dominated by less educated and unskilled manpower.
A considerable rise in the number of educational institutions and enrolment after 1980s is not yet reflected in Pakistans labour market. This might be due to the fact that most of the bachelors and masters degree programmes emphasise only on academic education without developing specific skills. The sluggish demand for the graduates of these programs in the job markets leads to unemployment among the educated and the job market remains dominated by the less educated. In this scenario, it becomes important to explore the role of education for the economic benefit of individuals.
6 III. Theoretical Model and Estimation Methodology: We start with the human capital model developed by Becker (1964) and Mincer (1974) where natural logarithm of monthly earnings are the linear function of completed school years, experience and its square. In mathematical form the equation can be written as: ln Wi = ? 0 + ? 1 EDU i + ? 2 EXPi + ? 3 ( EXPi ) 2 + Ui (1) where ln Wi stands for natural logarithm of monthly earnings, EDUi represents completed years of schooling, and EXPi is the labor market experience of ith individual.
?1 implies the marginal rate of return to schooling. A positive value of ? 2 and negative value of ? 3 reflects the concavity of the earning function with respect to experience. Ui is the error term, assumed to be normally and identically distributed. It has been argued in the literature that different school years impart different skills and hence affect earnings differently. Therefore, it is misleading to assume a uniform rate of return for all educational levels. Most of the previous studies used dummy variables to capture the effect of different levels of education.
In order to examine the effect of school years at different levels of education, van der Gaag and Vijverberg (1989) divided the years of schooling according to the school systems of Cote d Ivore. Similarly Khandker (1990) also used years of primary, secondary and post-secondary schooling in wage function for Peru. Both studies found significant differences in returns to education at different levels of education. Following van der Gaag and Vijverberg (1989), we divide the school years into seven categories according to the education system of Pakistan.
In Pakistan, the primary education consists of 5 years of schooling; middle requires 3 more years; and by completing 2 more years of schooling after middle, an individual obtains a secondary school certificate i. e. , Matric. After matric , i. e. , 10 years of schooling, students have a choice between technical and formal education. Technical education 7 can be obtained from technical institutions which award diploma after 3 years of education while the certificate of intermediate can be obtained after two years of formal education.
After the completion of intermediate certificate, students can enter either in the professional colleges for four years or in non-professional bachelors degree program for two years in a college. Those who choose non-professional degree can pursue their studies in a university for masters for two more years. At this stage the graduates of professional and non-professional colleges complete 16 years of education. They can now proceed to the M. Phil. or Ph. D. degrees. In order to examine the returns to education at different splines of education, we estimate the following extended earning function. ln Wi = ? 0 + ? 1Yrs Pr imi + ?
2 YrsMid i + ? 3YrsMati + ? 4 YrsInteri + ? 5 YrsBAi + (2) ? 6 Yrs Pr of i + ? 7 EXPi + ? 8 ( EXPi ) 2 + Ui where YrsPrim, YrsMid, YrsMat YrsInter YrsBA YrsProf are defined as: YrsPrim = D5EDUi YrsMid = D8EDUi YrsMat = D10EDUi YrsInter = D12EDUi YrsBA = D14EDUi YrsProf = D16EDUi where D5 = 1 if where D8 = 1 if where D10 = 1 if where D12 = 1 if where D14 = 1 if where D16 = 1 if 0< EDU ? 5 5< EDU ? 8 8< EDU ? 10 10< EDU ? 12 12< EDU ? 14 EDU > 14 The coefficients associated with YrsPrim, YrsMid, YrsMat YrsInter YrsBA YrsProf in equation 2 imply an increase in income with one year increase in education at respective levels.
For example, the returns to five completed years of education at primary level will be 5*? 1. Similarly, the returns to for six, seven and eight of education will be 5*? 1+? 2, 5*? 1+2? 2, and 5*? 1+3? 2 respectively. On the same lines we can compute the returns to education at each level as: 8 Returns to Primary =5*? 1 Returns to Middle =5*? 1+3*? 2 Returns to Matric= 5*? 1+3*? 2+2*? 3 Returns to Intermediate=5*? 1+3*? 2+2*? 3 +2*? 4 Returns to Bachelors =5*? 1+3*? 2+2*? 3 +2*? 4 +2*? 5 Returns to MA/Prof=5*? 1+3*? 2+2*? 3 +2*? 4 +2*? 5 +2*?
6 The data are drawn from the nationally representative Pakistan Integrated Household Survey 1995-96. In order to assess the performance of the Social Action Programme (SAP), the government of Pakistan has launched the series of Pakistan Integrated Household Surveys (PIHS), a collaborative nation wide data collection effort undertaken by the Federal Bureau of Statistics (FBS). So far two rounds have been completed. The first round of the PIHS is different from other round on two counts. Firstly, the information on employment and wages is available only in this round.
Secondly, only 33 percent of the sample used in the first round is being repeated in the subsequent rounds. This implies that all of these rounds are independent cross-section data sets and can not be properly linked with each other to be used as panel data. Therefore, the appropriate sample can only be drawn from the first round of PIHS. This round was conducted in 1995-96, which covers 12,622 households and more than 84,000 individuals. The 1995-96 PIHS provides a detailed information on completed school years5. In addition, this survey contains information on age started school.
This information is particularly important for our study to calculate the potential experience of a worker. The indicator for experience used by Mincer (1974) is a good proxy for U. S. workers as they start school at the uniform age of six years6. However, this assumption does not hold in Pakistan, as in this country there is no uniform age to start school. In urban areas, children as young as three years start going to school whereas in rural 5 This is the only nation-wide data set that provides this particular information.
Similarly no other survey contains information on public and private school attendence and year starting school. 6 Mincer defined experience as (Age-education-6). 9 areas the school starting age is higher. 7 This information enables us to construct potential experience as (age-schools years-age starting school). Although experience is still a proxy for actual experience but it is relatively better measure than age and the Mincer type potential experience. In addition to education and experience, various other factors, such as quality of schooling, technical training and quality of schooling have significant impact on earning8.
It has been argued that because of the market-oriented approach adopted by the private schools, the graduates of these schools earn more as compared to the graduates of public schools9. According to Sabot (1992), Behrman, Ross, Sabot and Tropp (1994), Alderman, Behrman, Ross and Sabot (1996a), Alderman, Behrman, Ross and Sabot (1996b), and Behrman, Khan, Ross and Sabot (1997), the quality of education has positive, significant and substantial impact on cognitive achievements and hence on post school productivity, measured by earnings.
These studies observed higher earnings of the graduates of high quality school than those who attended a low quality school. A recent study by Nasir (1999) found considerably higher earnings for the private school graduates. These schools, however, charge higher fees. Estimates of average annual expenditure per pupil in both government and private schools indicates that the total cost of primary level in rural areas is Rs. 437 (Rs 355 for government schools and Rs. 1252 for private schools), compared with Rs. 2038 in urban areas (Rs.
1315 for government and Rs. 3478 for private schools). This means that the cost of primary schooling is almost three times that of public schools in urban 7 The issue of age starting school has been highlighted by Ashraf and Ashraf (1993) and because of the nonavailability of this information, they used age as proxy for experience. 8 See Summers and Wolf (1977); Rizzuto and Wachtel (1980); Behrman and Birdsall (1983); Booissiere, Knight and Sabot (1985); Knight and Sabot (1990);Behrman, Ross, Sabot, and Tropp (1994); Behrman, Khan, Ross and Sabot (1997).
9 Various studies found the effectiveness of private schools to acquire cognitive skills [Colemen, Hoffer and Kilgore (1982); and Jimenez, Lockheed, Luna and Paqueo (1989)]. For Pakistan, Sabot (1992), Behrman, Ross, Sabot and Tropp (1994), Alderman, Behrman, Ross and Sabot (1996a), Alderman, Behrman, Ross and Sabot (1996b), and Behrman, Khan, Ross and Sabot (1997) found a significant variation in the cognitive skills among children with same number of school years. These studies conclude that some of the differences are due to the family characteristics while some are due to the quality of schooling. 10.
areas and nearly four times in rural areas. The differences in cost of schooling also reflect the degree of quality differentials in public and private schools, and between urban and rural schools. A relatively better provision of school facilities and quality of education in private schools is causing a continuous rise in school enrolment in urban areas [Mehmood (1999) page 20]. The PIHS provides information on the type of school attended10. On the basis of this information we can identify workers according to the school they attended and therefore examine the effect of type of school on individual earnings.
In order to capture the quality of education an individual received, a dummy variable is included in the model that takes the value 1 if individual is a graduate of private schools and 0 otherwise. The effect of post-school training on earning has been found positive and substantial in many developing countries [see Jimenez and Kugler (1987); van der Gaag and Vijverberg (1989); Khandker (1990); and Nasir (1999)]. The PIHS contains information on years of technical training. This information helps us to examine the effect of technical training received on individual earnings.
We use completed years of technical training as independent variable in the earning function. The existence of vast gender gap in human capital accumulation is evidenced by various studies in Pakistan11. The PIHS reports vast gender disparities in literacy and enrolment rates. The literacy rate among females is half than that of males literacy rate for whole Pakistan. This difference has increased to three-folds for rural areas. The gender difference is however smaller for the gross enrolment rate at primary level. For the higher levels of education, this difference 10.
The coefficient of private school may also capture the effect of socio-economic background of workers. The data, however, does not contain such information, therefore we are unable to separate the effect of parental characteristics from the effect of private schools in workers earnings. 11 Sabot (1992); and Alderman, Behrman, Ross and Sabot (1996b); Sawada (1997); Shabbir (1993); and Ashraf and Ashraf (1993a, 1993b, and 1996) 11 shows an increasing trend. Similarly vast gender gap has been observed in returns to education where males earn more than the female workers [Ashraf and Ashraf (1993a, 1993b and 1996) and Nasir (1999)].
In order to capture the effect of gender, a dummy variable is introduced in the model that takes the value 1 for males and 0 otherwise. The regional imbalances in the provision of limited available social services are more pronounced in Pakistan. Rural areas are not only underdeveloped in terms of physical infrastructure but also neglected in gaining basic amenities. Haq (1997) calculated the disaggregated human development index for Pakistan and its provinces. He noted that nearly 56 percent of population is deprived of basic amenities of life in Pakistan; 58 percent in rural areas and 48 percent in urban areas.
According to the 1995-96 PIHS, the literacy rate in urban areas is 57 percent and in rural areas it is 31 percent. The gross enrolment rate was noted 92 percent in urban areas and 68 percent in rural areas. Because of these differences low returns to education are observed in rural areas [Shabbir (1993 and 1994) and Nasir (1999)]. To capture the effect of regional differences, a dummy variable is used that takes the value 1 if individual lives in urban areas and zero otherwise. The four provinces of Pakistan exhibit different characteristics in terms of economic as well as social and cultural values.
Significant provincial differentials in rates of returns to education have been noted that reflect not only the differences in market opportunities but also indicate uneven expansion of social services across provinces [Khan and Irfan (1985); Shabbir and Khan (1991); Shabbir (1993); Shabbir (1994); and Haq (1997)]. The effects of these differences are captured through the use of dummy variables for each province in the earning function, Sindh being the excluded category. 12 For the purpose of analysis we restrict our sample to wage earners and salaried persons. Our sample contains 4828 individuals.
Among them, 4375 are males and 453 are females. Table 1 presents the descriptive statistics of some of the salient features of the important variables. According to the statistics in table 1, average age of the individuals included in the sample is 34 years with 18 years of experience. A typical worker in the sample has completed approximately 10 years of education. A majority is graduated from public schools. Most of the workers live in urban areas. On average an individual earns Rs. 3163 per month. In our sample, there are only 22 percent individuals who received technical training.
The average years spent for training are less than one year. A majority of wage earners belong to Punjab, followed by Sindh and Balochistan. Table1 Mean, Standard Deviation and Brief Definitions of Important Variables Variables W Age EDU EXP RWA MALE Urban Private Training Punjab Sindh NWFP Balochistan Mean SD Variables Definitions 3163. 34 3397. 39 Individuals monthly earnings in rupees consist of wages and salaries. 34. 07 12. 36 Age of an individual in years. 9. 53 4. 36 Completed years of schooling. 18. 14 11. 80 Total Years of labour market experience calculated as (age-school years-age starting school).
2. 37 1. 07 Categorical variables, contains 4 categories of literacy and numeracy. 0. 91 0. 29 Dichotomous variable equal to 1 if individual is male. 0. 60 0. 49 Dichotomous variable equal to 1 if individual belongs to urban area 0. 04 0. 19 Dichotomous variable equal to 1 if individual is a graduate of private school 0. 35 0. 87 Completed years of technical training 0. 38 0. 49 Dichotomous variable equal to 1 if individual belongs to Punjab 0. 31 0. 46 Dichotomous variable equal to 1 if individual belongs to Sindh 0. 15 0. 36 Dichotomous variable equal to 1 if individual belongs to NWFP 0. 16 0.
36 Dichotomous variable equal to 1 if individual belongs to Balochistan 13 IV. Empirical Results The estimated results of equation 1 and equation 2 are reported in table 2. The highly significant coefficients of school years and experience indicate the applicability of human capital model for Pakistan. An additional year of schooling raises individuals monthly income by 7. 3 percent, which is very close to the prior studies. 12 13 The coefficient of experience shows substantial increase in wages with each additional year. The concavity of age-earnings profile is evident from the negative and significant coefficient of experience squared.
The results reveal that an individual with five years of experience earns 31 percent higher wages as compared to non-experience worker. The highest level of earnings is achieved with approximately 30 years of experience. These estimates are relatively low compared to prior studies14. The positive and significant coefficients of gender (0. 401) and regional dummies (0. 178) strengthens the a priori expectation that males earn more than females and earnings are higher in urban areas as compared to rural areas. These estimates are consistent with earlier studies [see Arshaf and Ashraf (1993), Khan and Irfan (1985)].
Furthermore, significant inter-provincial differences in individuals earnings can be observed in the estimated model. Many studies indicate substantial differences in earnings across school levels. For example, van der Gaag and Vijverberg (1989) noted that an increase of one year in elementary, high and university education causes an increase of 12 percent, 20 percent and 22 percent respectively in 12 The estimated coefficients of school years by Shabbir and Khan (1991), Shabbir (1991), Shabbir (1993) and Shabbir (1994) are found to be in the range of 6 percent to 9.
7 percent. 13 The returns to education are calculated by taking the anti-log of 0. 092 (estimated coefficient of completed school years) and subtracting from 1. To convert into percentage, multiply the value by 100. For details, please see Gujrati (1988) page 149. 14 The difference in the returns to experience could be due to the approach adopted by these studies. Most of the studies used age as a proxy for experience [see for example Khan and Irfan (1985); Ashraf and Ashraf (1993); and Nasir (1999)]. Shabbir (1991) used the Mincerian approach to calculate experience.
The present study uses actual age of starting school and actual years of education. These information enable us to calculate total years of labor market experience. This approach is also not the perfect alternative for actual experience, as we do not have information about the starting time of the first job. But when compared with other approaches, it is more precise in measuring experience. 14 earnings. In order to examine the returns to education across different school years, we include the information on schooling according to the education system of Pakistan (equation 2).
The results reported in column 3 of table 2 show a positive and significant impact of school years at each educational level on earnings. For example, an increase of one year in education at primary level increases the earnings by 3 percent. Similarly, at middle level, one year of schooling brings about an increase of 4 percent in earnings and the total returns to schooling at middle level are 27 percent. Table 2 Earning Function with and without Levels of Education Variables Coefficient s 6. 122 0. 072* 0. 058* -0. 001* 0. 178* 0. 401* 0. 127* -0. 113* -0. 203* 0. 412 t-ratios Coefficient s 6. 380 0. 058* -0.
001* 0. 150* 0. 264* 0. 098* -0. 112* -0. 166* 0. 027** 0. 040* 0. 050* 0. 057* 0. 071* 0. 082* 0. 429 t-ratios Coefficient s 6. 342 0. 058* -0. 001* 0. 152* 0. 262* 0. 096* -0. 108* -0. 164* 0. 052* 0. 007 0. 025* 0. 038* 0. 047* 0. 063* 0. 075* 0. 429 t-ratios Constant EDU EXP EXP2 Urban Male Balochistan NWFP Punjab RWA Yrs-Prim Yrs-Mid Yrs-Mat Yrs-Inter Yrs-BA Yrs-Prof Adj R2 148. 91 46. 71 26. 49 -19. 20 10. 31 13. 98 4. 94 -4. 34 -10. 21 92. 03 23. 85 -16. 84 7. 87 8. 15 3. 40 -4. 06 -7. 75 2. 03 5. 07 8. 69 11. 41 16. 85 21. 98 89. 25 23. 84 -16. 88 7. 98 8. 09 3. 32 -3. 91 -7. 63 2. 41 0. 45 2.
45 5. 02 7. 28 11. 47 15. 57 * significant at 99 percent level. ** significant at 95 percent level. One can note higher returns of additional year of schooling for higher educational levels from this table. For example, the returns to masters and professional education (Yrs-Prof) are more than five- 15 times higher than that of primary school years (Yrs-Prim). The results exhibit a difference of 15 percent between primary graduates and illiterates, the excluded category. This category includes illiterates as well as all those who have not obtained any formal schooling but have literacy and numeracy skills15.
To further explore the earning differential between primary school graduates and those who never attended school but have literacy and numeracy skills, we have constructed an index RWA that separates illiterates from those who have literacy and numeracy skills. This index takes the value zero if individual does not have any skill; 1 if individual has only one skill; 2 if individual has two skills; and 3 if individual has all three skills. We re-estimated equation 2 with this new variable and the results are reported in column 5 of table 2.
According to our expectations, the coefficient of RWA is found not only large (0. 05) in magnitude but also statistically significant at 99 percent level. This indicates that the individuals with all three skills earn 15 percent more than those who have no skill. On the other hand, the coefficient of Yrs-Prim dropped to 0. 007 and became insignificant16. The differential in the earnings of illiterates and those having five years of primary education was 15 percent (0. 03*5=0. 15). This differential however, reduced to approximately 9 percent (0. 007*5+0. 053=8.
8) when we include those who have no formal education but have literacy and numeracy skills. These high returns to cognitive skills indicates the willingness of employer to pay higher wages to the able workers as compared to those who have five or less years of schooling but do not have these skills. Now we examine the effect of technical training and quality of schooling on earnings, first in separate equations and then in a single equation. The impact of technical training on earnings is examined by including years of apprenticeship as continuous variable in our model.
The results are reported in column 1 of table 3. The results show a positive and significant impact of technical 15 There are 48 wage earners in our sample who have education less than primary but do not have any of these skill. Whereas we found 76 wage earners who do not have any formal education but have at least one of these skills. 16 This result is consistent with van der Gaag and Vijierberg (1989). 16 Table 3 Earning Functions : Impact of Technical Training and School Quality (Separate Functions) Variables Constant EDU EXP EXP2 Urban Male Balochistan NWFP Punjab Train.