1. First World KLEMS conference
India KLEMS
Labour Input- Quantity and Quality by Industry
Suresh Aggarwal
First World KLEMS conference
Harvard University
19-20 August 2010
Research assistance by Gunajit Kalita in creating the India KLEMS Labour Input dataset 1

2. Objectives-India KLEMS
To create a comprehensive data base on productivity growth using Growth Accounting Approach.
Construct a Time Series data on output, capital, labour, labour quality and intermediate inputs. 2

3. Major tasks for Data Base on Labour
Make a Time series of Employment from 1980 to 2004.
Prepare a Labour Quality Index from 1980 to 2004.
Make a Time series of Labour Input from 1980 to 2004. 3

4. Major Contributions of the Paper
Efforts have been made for the first time to estimate employment in Hours.
Average number of Hours worked in a day have been estimated for the first time.
Both the Quinquennial and the annual rounds have been used, for the first time for constructing the time series of employment.
A separate decomposition of Labour Quality into indices of age, sex and education has been attempted.

5. Broad classifications for all the series
Gender: Males/Females
Age : <29; 30-49; and 50+
Education: Up to Primary; From Primary to Higher Secondary; and above Higher Secondary.
Sectors : 31 sectors.
So it is 2*3*3*31 classification. 5

6. Major Sources of Data Used
For all sectors of the economy Employment and Unemployment Surveys (EUS) by National Sample Survey Organization (NSSO) and Population Census. The two are Household/Individual specific.
Manufacturing Sector:
Organized Manufacturing industries-Annual Survey of Industries(ASI) by Central Statistical Organization (CSO).
Unorganized Manufacturing industries- Residual. 6

7. Methodology for Constructing the Time Series of Employment
Time Series of employment requires estimation of: a) Number of persons, and
b) Total days and hours worked by each person.
Time Series of Labour Input- Number of persons employed
In India, the number of employed may be estimated from Census and/or from EUS.
While Census has been held every ten years, NSSO has conducted both major (or Quinquennial) and thin (or annual) rounds of EUS. 7

8. Employment Unemployment Survey (EUS)
Major (Quinquennial) Rounds of EUS since 1980: 38th (1983), 43rd( ), 50th( ),
55th ( ) and 61st( ).
Thin (Annual) Rounds: 45th to 60th .
EUS uses Usual Status [Usual Principal Status(UPS) and Usual Principal & Subsidiary Status (UPSS)], Current Weekly Status(CWS) and Current Daily Status (CDS) measures for Quinquennial (or major) rounds and Usual Status & CWS for annual (thin) rounds.
While UPS, UPSS and CWS measure number of persons, the CDS gives number of jobs. 8

9. Definition of UPSS, etc.
The usual principal status gives the number of persons who worked for a relatively longer part of the reference period of 365 days preceding the date of survey,
While the usual principal status and the subsidiary status, includes the persons who (a) either worked for a relatively longer part of the 365 days preceding the date of survey or (b) who had worked some time (minimum 30 days since 61st round) during the reference period of 365 days preceding the date of survey.
The current weekly status provides the number of persons worked for at least 1 hour on any day during the 7 days preceding the date of survey, and
The current daily status gives the picture of the person-days worked during the reference week of the survey period.

10. contd….
UPSS is the most liberal and widely used of these concepts. It includes all workers who have worked for a longer time of the preceding 365 days in either the principal or in one or more subsidiary economic activity.
Advantages of using UPSS
It provides more consistent and long term trend.
More comparable over the different EUS rounds.
When adjusted for population distribution, it provides the count of jobs.
Wider agreement on its use for measuring employment [Visaria(1996), Bosworth; Collins & Virmani (2007), Sundaram (2008), Rangarajan (2009)].
It can also be calculated for thin rounds. 10

11. For India KLEMS we have used UPSS to estimate employment.
contd….
Some problems in using UPSS
Seeks to place as many persons as possible under the employed;
No single long-term activity status for many due to movement to many jobs.
Requires a recall of one year.
Though, UPSS has some limitations, but this is the best measure to use given the data.
For India KLEMS we have used UPSS to estimate employment. 11

12. EUS- contd…. For India KLEMS we have used UPSS to estimate employment.
Both the Quinquennial and the annual rounds have been used, for constructing the time series.
Since different rounds of EUS use different National Industrial Classification (NIC), so a Concordance between India KLEMS, NIC-1970, 1987 and 1998 required for all the 31 sectors has been done.
‘Total hours worked’ have been estimated by also using the CDS schedule of the EUS.

13. Time Series of Labour Input- Numbers
But if only major rounds of EUS are used for estimating Employment, then we have data only on selected five points.
So the issue was of constructing a time series from these data points.
Alternatives were:
Interpolation from These Five Points.
Since early nineties, annual (thin) round data is also available. Combine it with major rounds. 13

14. Time Series of Labour Input –Numbers
If thin rounds are also used, then the issues are:
Comparing major rounds with thin rounds; and
Obtaining three digits data through thin rounds.
Accepting the suggestions of the experts, we made use of both the major and the thin rounds. 14

15. Contd…
Since different rounds of EUS use different NIC, so a Concordance between India KLEMS, NIC-1970, 1987 and 1998 required for all the 31 sectors was done.
The interpolation from the major rounds was done for the period to The interpolated numbers were then constrained by the numbers obtained from the industrial distribution of the thin rounds.

16. Total hours worked
Once the numbers were obtained ,efforts were made to obtain ‘total hours worked’ estimates from:
EUS - Time Disposition during the week
CDS and use intensity of work – Full time( ≥4hours) and Part time (<4hours).
Information on man-days workers and man-days employees at all India for all manufacturing industries was taken from ASI.

17. Estimation of Employment
Employment has been computed as follows:
Used; like all the previous studies, the Work Participation Rates (WPRs) by UPSS from EUS and applied them to the corresponding period’s population of Rural Male, Rural Female, Urban Male and Urban Female to find out the number of workers in the four segments .
Use the 31-industry distribution of Employment from EUS and used these to the number of workers in step I and obtained Lij for each industry where i=1 for rural and 2 for urban sectors, and j=1 for male and 2 for female.

18. Contd….
III. Find out the average number of days worked per week ‘dij’ for each industry from the intensity of employment as given in the CDS schedule. IV. Assuming average 48 hours work week for regular workers and 8 hours per day for self employed and casual workers, find out the expected number of hours ‘hij’ worked per day from the status-wise distribution, in each industry for rural male, rural female, urban male and urban female.

19. Contd….
From the major rounds separate interpolation of Lij ; dij; and hij was done for rural male, rural female, urban male and urban female to obtain the respective time series.
Broad Industrial distribution from annual rounds was used as a control total on the corresponding interpolated Lij and revised numbers were obtained.
Total person hours in a year were obtained for each industry as the sum of the products of revised Lij; dij; and hij over gender and sectors. ΣiΣjLij*dij*hij*52

20. Time Series of Labour Quality Index
Quality Index has been constructed using the standard methodology given by Jorgenson, et al (1987), which uses the Tornqvist translog index.
Analogously, other first order contributions by gender, age and education, Qs , Qa, and Qe , have also been computed.
Data required for Quality Index is:
Since the required labour composition data is available only from major rounds of EUS, so
Only Major rounds have been used for estimating the indices and the indices have been interpolated to get the time series for the entire period.
Only for aggregate 31 sectors- not for organized and unorganized separately.
Employment by sex by age by education by industry;
Earnings for each of these cells.

21. The growth rate of the quality index QL can be expressed in the form:
They have expressed the volume of labour input, L; as a translog index of its individual components and the weights are given by the average shares of the components in the value of labour compensation.
The growth rate of the quality index QL can be expressed in the form:
ln QL = Σlvll ln Ll - ln L
where L= ΣlLl
QL is the quality index of labour, and L is the total number of labour (unadjusted) of all education categories.
This is thus, the difference between the percentage change in quality- adjusted labour and the percentage change in actual labour, summed over all categories.

22. Time Series of Labour Quality Index
Quality Index has been constructed using Jorgenson, et al (1987) methodology which uses the Tornqvist translog index.
They have expressed the volume of labour input, L; as a translog index of its individual components and the weights are given by the average shares of the components in the value of labour compensation.
The growth rate of the aggregate labour volume index is defined as:
where Lw is the weight adjusted aggregate labour,
Ll is labour of a particular education class,
l= 1,2,…..,n i.e. the number of education categories,
vl is the value share of labour for the lth education category,
is the wage rate of labour for the lth education category,
is the summation over all education categories.
and 22

23. Contd…
Growth of labour volume L incorporates both growth in hours worked and improvement in labour quality.
Since data on hours worked for each educational category of labour is not easily available, we assume that labour input for each category is proportional to hours worked and the proportion is same for all categories.
It follows from this that the growth rate of the quality index QL can be expressed in the form:

24. Earnings Data
NSSO’s EUS relates earnings to only regular- salaried workers and casual workers.
The issue was how to estimate earnings of self employed.
The present study has used the Mincer Wage equation for the same and sample selection bias has been corrected for by using Heckman's two step procedure.
Earnings of Self Employed is required for quality index and labour compensation.

25. Earnings of Self Employed by KLEMS
OECD assumes that labour characteristics of both employees and self employed is same within an industry. So average compensation per hour of a self employed person is taken to be equal to that of a wage earner.
EU Klems has followed the OCED procedure for most of EU countries, but on the basis of some surveys in few places they have estimated it to be 0.80 for some sectors, especially agriculture and 1.20 for sectors like business services. 25

26. Earnings of Self Employed in India
Two alternatives were considered :
Use earnings of Self employed to be equal to that of Casual labour as the labour market for the two is comparable.
To fit an earning function to earnings of casual and regular employees and use it to find the corresponding earnings of the self employed.
India KLEMS preferred to use the second option and has used Heckman's procedure for the same

27. Results Results are presented as follows:
Firstly, for the Total economy.
Secondly, by the broad industrial classification.
Lastly, by the 31 KLEMS industrial classification.

28. Results- Total Economy
Profile of Worker’s Population in India
Only marginal changes in WFPR between 38th and 61st round; at around 42% with a tendency to increase in urban sector and reduce in rural sector.
WFPR are higher for males than for females, and for rural females than urban females (it is 1/4th to 1/3rd for rural females and 1/7th to 1/6th for urban females).

29. Workforce Participation rate in different NSSO rounds (% of Total Population)

30. Labour Input and Quality Change for the Total Economy

31. Growth Rates of Labour Input, Hours and Labour Quality (% per annum)
1980 to
1985
1986
1990
1992
1996
1997
2004
1989
1999
2001
GDP
5.28
5.89
6.54
5.93
5.71
5.58
6.16
6.41
Variable Labour
Labour Input
1.82
2.93
2.49
2.64
2.01
2.46
3.42
Labour Persons
1.20
1.55
1.66
2.15
1.85
1.15
1.64
2.83
Labour Hours
1.46
2.55
2.10
2.14
2.22
1.65
2.06
2.85
Labour Quality
0.35
0.37
0.39
0.50
0.41
0.36
0.56
First order
Quality Indices
Qs (Gender)
0.01
-0.01
0.00
-0.02
Qa (Age)
0.07
0.06
0.04
0.03
Qe (Education)
0.28
0.33
0.48
0.38
0.30

32. Total Employment (persons and million hours) and hours per day

33. Results Over the period of 1980-2004:
The growth in labour quality is 0.41 % per year.
The growth in employment in hours is faster than growth in employment by persons.
There is just a marginal growth in days per week.
Hours per day have almost remained constant.
While the labour quality growth is highest in the most recent period of 1997 to 2004 at 0.50%; the growth in labour input is highest in the period of 1986 to 1990 at 2.93%.

34. Contd….
However, decade-wise periodization shows that both labour input and labour quality have experienced the highest increase in the current decade, i.e. from 2001 to when the Indian economy is booming and the GDP growth is highest.
The contribution of education is 0.38 percentage points out of 0.41 percentage points of labour quality during the period.
The decade-wise analysis indicates that in the recent period the entire growth in labour quality is contributed by education indices and this contribution has increased over the previous two decades.

35. Aggregate Quality and its first order Approximation

36. Comparison with two other major studies
Author
Period
Growth rate in Employment Index
Growth in Education Index
Growth in Labour Input Index
Bosworth; Collins & Virmani (2007)
2.00
0.40 -
Sivasubramonian (2004)
1980 to 1999
1.74
0.34
2.22
1980 to 1990
2.02
0.31
2.47
1990 to 1999
1.43
0.37
1.93
Current study (2010)
1980 to 2004
1.85
0.38
2.64
1980 to 1989
1.15
0.30
2.01
1990 to 1999*
1.64
0.35
2.46
*Year 1991 has been excluded from the current study because of it being an abnormal year
The results for employment growth are different from Sivasubramonian’s study, but are close with Bosworth; Collins & Virmani (BCV).
The results for education growth rates are however, very close.

37. Composition of Labour Education
The proportion of more educated workers has increased, and of literate up to primary has reduced
Cumulative Distribution of educational attainment of workers
Above Higher Secondary
Primary to Higher Secondary
Upto Primary

38. Education-contd…
In year 2004, while a worker with education above higher secondary was getting 2.5 times the wages of a worker with education from primary to higher secondary, a worker with education up to primary only was getting just two-third of it .
The wage differential has increased for a worker with above higher secondary education over the period.
The differential reduced however, between the first two education categories from less than half to two-third now.
The average number of days worked by workers has consistently increased over the years from 5.24 days to 5.68 days at an average annual growth rate of 0.39 %.
Along with an increase in the number of persons, the increase in the number of days has contributed to the increase in labour input.
The increase in days is maximum for the most educated category and minimum for the least educated category.

39. Relative Wages of workers by Educational Attainment

40. Distribution of workers by age groups
The proportion of younger age group <29 has declined and the proportion of middle age (30-49) has increased; with no change in 50+.
Distribution of workers by age groups
<29
30-49
50+

41. Age-contd…
It is the middle age group (30-49) which has been getting the highest compensation .
The relative wages have however, generally improved for the remaining two age groups over the period .
The average number of days worked by workers of age group has been consistently more than the other two age groups .
Relative Wages of workers by age
30-49
<29

42. Gender
The proportion of female workforce has almost remained stagnant over the time period at 26%.
The females are also catching up in terms of number of days per week and number of hours per day worked.
The catching up is showing in the trend in relative wages which have marginally increased from 62 per cent to two-third.
The gender quality index for the period remained stagnant .

43. Gender: Female’s share of workforce, relative wages, days and hours
Days Per Week Ratio
Wage Ratio
Share of Workforce

44. Employment Class
In India the employment class is broadly divided in to self employed, regular employees and casual employees.
Self employed constitute more than half of the total employment in India. This proportion declined till 1999 but has again increased in the recent years .
Except for the last round, while the share of regular employees has remained constant that of casual labour has increased.
The regular employees work for the maximum days (6.77 days in 2004) in a week and the casual employees work for the minimum days( 5.09 days).

45. Percentage Distribution of Employment by Type

46. Labour Input by the broad industrial classification.
Industry
1980 to 1985
1986 to 1990
1992 to 1996
1997 to 2004
1980 to 2004
1980 to 1989
1990 to 1999*
2001 to 2004
Agriculture
0.63
1.19
1.46
0.84
1.37
0.79
1.08
1.25
4.29
3.52
2.44
4.84
3.88
3.76
2.45
6.80
Services
3.00
6.70
5.35
4.27
4.45
3.68
5.76
4.88
Total Economy
1.82
2.93
2.49
2.64
2.01
2.46
3.42
* Excludes 1991

47. cont.…
The growth in labour input is driven by the service sector over the period
Agriculture has been the laggard in growth of employment.
While manufacturing produced faster employment growth in 1980’s, services led the growth in 1990’s.
In the recent decade, manufacturing has again taken the lead in faster growth of employment.
The growth in employment has been faster in the second decade of reforms contrary to the belief of many.

48. Distribution of Workers by Industry: Labour Input Levels
The biggest industry for labour absorption is agriculture, forestry and fishery; followed by retail trade (industry 20); construction (industry 17) and transport & storage (industry 22).
The high labour input industries, where more than 80% of the output receipt is paid to labour are public administration and Defence (industry 27), Construction (industry 17), agriculture (industry 1) and other Community, Social and Personal Services (industry 30).
At the other extreme are very low labour intensive industries, ex. Coke, Refined Petroleum and Nuclear Fuel (industry 7) where the labour share is only 4.4 percent.

49. Labour Characteristics by Industry in 2004
Across all 31 industries, the median for workers with above Higher Secondary education is %.
Ranking of industries indicates that the industries with the highest proportion of above Higher Secondary educated workers in 2004 are education, financial intermediation, renting of machinery, health and social work, and Coke, refined petroleum and nuclear fuel .
The industries with very low (less than 5 %) proportion of more educated labour are Private households with employed persons; agriculture; wood and products of wood; construction; other non-metallic minerals; food and beverages and tobacco; hotels and restaurants; and textiles, textile products, etc.

50. contd…
The industries that have higher proportion of workers with above Higher Secondary education are generally also the ones with higher compensation of labour and the vice versa .
The average share of the female hours worked is about 18.1 per cent .
The female’s share is very low in Transport and storage; Sale, maintenance & repair of motor vehicles; Transport equipment; Wholesale trade; Real estate and Construction with just 10 per cent share.
The industries with high share are Private households with employed persons (71%) and Food & beverages & tobacco (44%).

51. Labour Input Growth rates in Industry
The overall proportion of males in total employment is almost 2/3rd in all the rounds while it is just 1/3rd for females.
The proportion of female workers is more than males in private households with employed persons and is expectedly quite high in few other industrial groups e.g. in education, health and social work, food & beverages and in agriculture, forestry& fishing.
Though the dependence on agriculture for employment has reduced, still agriculture remains the biggest employer of workforce and employs 54 per cent of them in 2004.
The other industries which are the major source of employment for males are retail trade; construction; transport & storage; and textiles, and for female workers it is mainly education; textiles; food & beverages and trade.

52. Contd…
There have been structural changes in terms of employment growth in the Indian economy. While some industries grew very fast, the others remained stagnant during this period.
The industries which grew the fastest have been real estate activities; renting of machinery; construction; post & telecommunication; sale, rubber & plastics; maintenance and repair of motor vehicles; and financial intermediation all growing at more than around 6 per cent.
The industries with slow employment growth has been agriculture; wood and products of wood; public administration; food & beverages and textiles.

53. Growth in Labour Hours and Labour Input by Industries

54. Growth and Acceleration in Labour Quality
Growth in Quality
Acceleration in Quality

55. Labour Quality in Industries
The growth in labour quality was fastest in real estate activities; machinery; electricity, gas & water supply; and financial intermediation and very slow in wood & products of wood; construction; non-metallic minerals, agriculture and wholesale trade & commission.
The growth in labour quality was only 0.19 per cent in the pre reform period and it increased to 0.29 in the post reform decade indicating change in the composition of the workforce.

56. Contd….
The inter industry differences in the pattern of change in growth rate shows that the variation in growth rates has reduced over the period
The industries with either negative or very low growth rate in the first sub period (Sale, maintenance of motor vehicles etc., Construction, mining & quarrying, etc.) have generally been able to pick up the growth rate in the last period.
The reverse has also happened where the growth rate in labour quality for these industries has slowed down over the period (real estate, chemicals & chemical products, financial intermediation, etc.).

57. Bottom Five Industries
Labour Input
Top Five Industries
Bottom Five Industries
Growth in Labour Input (% per year), 1980 to 2004
Real Estate Activities
Agriculture, Forestry And Fishing
Renting Of Machinery And Equipment And Other Business Activities
Wood And Of Wood And Cork
Construction
Public Admin And Defence; Compulsory Social Security
Post And Telecommunications
Food And Beverages And Tobacco
Rubber & Plastic
Textiles, Textile Products and Leather
Change in Labour Input Growth, 1997 to 2004 less 1980 to 1985
Private Households With
Employed Persons
Coke, Refined Petroleum And Nuclear Fuel
Rubber And Plastics
Health & Social Work
Electrical & Optical Equipment

58. Bottom Five Industries
Labour Quality
Top Five Industries
Bottom Five Industries
Growth in Labour Quality (% per year), 1980 to 2004
Real Estate Activities
Private Households With Employed Persons
Machinery, Nec
Wood And Of Wood And Cork
Electricity Gas And Water Supply
Construction
Mining And Quarrying
Other Non-Metallic Mineral
Financial Intermediation
Agriculture, Forestry And Fishing
Change in Labour Quality Growth, 1997 to 2004 less 1980 to 1985
Sale, Maintenance And Repair Of Motor Vehicles And Motorcycles;
Retail Sale Of Fuel
Chemicals And Chemical Products
Post And Telecommunications
Private Households With
Employed Persons
Education

59. Distribution of Manufacturing Workers into Organized and Unorganized Sectors
The share of unorganized sector in the Indian manufacturing has consistently increased from 78 percent to 85 percent over the period from 1983 to So more and more employment is being sought in the unorganized sector.
While few industries, e.g. Wood and products of wood and cork; Textiles, Textile products & Leather & footwear; Food & beverages & tobacco; other non-metallic mineral; and Manufacturing nec., are highly concentrated in the unorganized sector, there is other extreme of coke & Petroleum where most of employment is in the organized sector.

60. Manufacturing Employment- Organized & Unorganized Sector

61. Conclusion The WFPR remained almost unchanged over the period.
The share of age-group is highest.
The share of educated workforce has gradually increased during the period.
There is a tendency for the share of female workers to increase, though the share is still less than half to that of males.
Nominal Wages are generally higher for more educated and experienced workers.
Along with increase in employment of labour hours there has also been increase in labour quality, leading to a faster growth of labour input.
The share of unorganized employment has increased in the Indian manufacturing sector.

62. Thank You 62