Reproduced from my article in the Economic and Political Weekly May 30, 2015
Unemployment rates in
India are not particularly interesting. Available most frequently from the
National Sample Survey (NSS), they are relatively low and have barely changed over
the years.
Recently available census data provides far higher estimates of unemployment than the NSS.[1] Census unemployment numbers are especially high for youth aged 15 to 29 who are new entrants to the labour force. Over forty million youth are seeking work and the unemployment rate stands at above 15% and 30% for young men and women respectively.[2]These numbers make the case for a closer look at census and NSS employment and unemployment data to understand why they differ and what they represent.
This article begins with
a brief discussion of the census and NSS methodologies to identify comparable
data from the two data sets. This is followed by a comparison of aggregated
employment, underemployment and unemployment data from the two sources to see
how well they agree and how that agreement has changed over time. The last
section looks at disaggregated data from the last census for clues to explain
the wide difference in unemployment estimates.
Areas of broad compatibility
For estimates from two
different data collection exercises to be comparable, they must at a minimum
work with similar concepts and have the same reference period[3].
The areas of broad compatibility between
the NSS and the census starting with the 1991 census are briefly captured below.[4]
Both the NSS and the census divide the country into rural and urban areas for data collection in a mutually consistent manner.[5] The census uses a reference period of one year for employment data and the NSS also provides data for such a period.
The census provides
only limited data on employment and unemployment. It separates the population
into ‘workers’ and ‘non-workers’ and provides estimates of the number of
‘workers’, the extent of underemployment among ‘workers’ and unemployment among
‘non-workers’. These then are the data of interest for a comparison between the
census and NSS.
Both the census and the
NSS have similar criteria for identifying ‘workers’ and the ‘unemployed’. ‘Workers’
are persons engaging in ‘economic activity’ and the set of production related
activities accepted as ‘economic activity’ is almost the same in both exercises.
The ‘unemployed’ among the non-working population are identified by determining
if they are ‘seeking work or available for work if there is work available’. In
urban areas, this test requires a person to be making tangible efforts to
obtain work. In rural areas it is less stringent and only requires that a
person be available for work if work is available. Both exercises estimate
underemployment in workers and both identify students in the non-working
population.
Getting to the
specifics, Census B Series tables provide the count of ‘main workers’ - workers
who have worked six months or more in the last year and ‘marginal workers’ –
those who have worked less than six months.[6]
The broadest count of workers is obtained by adding the count of ‘main workers’
and the count of ‘marginal workers’. From among the rest termed the ‘non-workers’,
those who pass the test of ‘seeking or available for work’ make up the count of
the census unemployed.[7]
The census also measures underemployment among ‘marginal workers’ by applying
the above test. Non-workers are additionally categorised by ‘main activity’
including as ‘students’ and persons ‘engaged in household duties’.
In the NSS, a person’s
‘activity status’ can be ‘worker’, ‘unemployed’ or ‘not in the labour force’
during the reference period and the ‘principal status’ is determined by the ‘activity status’ in which relatively
longer time has been spent (also referred to as the ‘major time’ criteria). [8]
The broadest count of workers with a 1 year reference period is obtained from
the tabulation referred to as ‘usual activity status (principal status +
subsidiary status)’ in the NSS “Employment and Unemployment Situation” reports.[9]
This includes persons whose ‘principal activity status’ is ‘worker’ as well as
persons whose ‘principal activity status’ is other than ‘worker’, but who have
worked for at least 30 days in the reference period. This tabulation also
provides the categorization of non-working persons as ‘unemployed’, ‘students’
and as those ‘engaged in domestic duties’. The NSS additionally provides
several indicators of underemployment among ‘workers’ which will be discussed
later.
Worker population ratios
Large sample surveys
are conducted every five years and the NSS large sample rounds closest to
census years have been selected for the comparison. Table 1 compares NSS and
census worker population ratios across 3 censuses. The female worker population
ratios in the NSS (1993-94) round show a large difference with the
corresponding ratios of Census 1991.
Table 1. Workers in every 1000 persons aged 15 and above
Men
|
Women
|
|||
Rural
|
Urban
|
Rural
|
Urban
|
|
Census - 1991
|
825
|
726
|
406
|
136
|
NSS (1993-94)
|
864
|
768
|
486
|
223
|
Census - 2001
|
810
|
720
|
463
|
167
|
NSS (1999-00)
|
841
|
752
|
452
|
197
|
Census - 2011
|
779
|
724
|
430
|
202
|
NSS (2011-12)
|
800
|
741
|
352
|
195
|
A government committee
of experts that was privy to the 1991 census data attributed the differences in
census and NSS estimates to two main factors – a) differences in the set of activities
that constituted ‘economic activity’ and b) differences in the minimum time
spent working to be considered a ‘worker’.[12]
More specifically, the
National Statistical Commission (also known as the Dr Rangarajan Commission)
was of the view that the exclusion of certain non-market
economic activities from the definition of work adopted in the census could be
the reason for the low female worker population ratios of Census 1991 in
comparison with the NSS 50th round.[13]
The scope of non-market economic activities was expanded in Census 2001 by
treating persons engaged in cultivation and growing of crops solely for
domestic consumption and persons engaged in rearing of animals for production
of milk for their own use as ‘workers’.[14]
There is a dramatic improvement seen in the agreement between Census 2001 and
NSS female workforce ratios, possibly as a consequence of this change. The
differences in the male workforce ratios have also narrowed further compared to
Census 1991.
Census 2011 shows a further narrowing of differences with NSS workforce
estimates for rural men and urban men and women. The only spoiler is the count
of rural working women where there is a large difference.[15]
The sharp fall in rural
female workforce ratios seen in both the NSS 66th round (2009-10)
and the NSS 68th round (2011-12) compared to NSS 61st
round (2004-05) without any accompanying increase in unemployment has puzzled researchers
following employment trends. There is an extensive literature attempting to throw
light on this phenomenon termed the “missing labour force”.[16]
If Census data is to be believed, much of the rural female labour force may not
have gone missing at all.
The next two sections
feature comparisons of underemployment and unemployment estimates. The
comparisons will be confined to Census 2001 and Census 2011 which have better
comparability with NSS data than Census 1991 for reasons stated earlier.
Underemployment in workers
The census, as
mentioned earlier, counts the number of ‘marginal workers’ who are seeking or
available for work.This is a measure of underemployment in workers who have
worked for less than six months.
The NSS has several indicators of underemployment among usual workers.
Among them, the one with a liberal interpretation of underemployment, is the
one which measures usual workers (principal status+ subsidiary status) who were
without work for at least 1 month and who sought or were available for work on
at least some days during those month(s). Table 2 shows the underemployment numbers from the census and the range
provided by this NSS indicator.
Table 2. Underemployed workers in every 1000 persons aged 15
and above
Men
|
Women
|
|||
Rural
|
Urban
|
Rural
|
Urban
|
|
Census - 2001
|
52
|
25
|
36
|
9
|
NSS (1999-00)
|
183-203
|
67-75
|
93-112
|
19-25
|
Census - 2011
|
98
|
32
|
69
|
17
|
NSS (2011-12)
|
155-160
|
48-53
|
58-62
|
15-16
|
Source: Census of India
and NSS ‘Employment and Unemployment Situation in India’ reports[17]
As the table shows, the
census generally underestimates underemployment when compared with this NSS
indicator. The probable reason for the wide margin of underestimation is that
the census does not test ‘main workers’ – those who have worked greater than
six months in the year - for underemployment. This means that the census
population for the underemployment test is far smaller than the NSS population
which includes all workers.
Unemployment and underemployment in non-workers
Table 3 compares census
figures of non-working persons ‘seeking work’ with the NSS numbers of ‘unemployed’
for the last two censuses. NSS figures show dramatically lower unemployment
than the census.
Table 3. Non-workers seeking work in every 1000 persons aged
15 and above
Men
|
Women
|
|||
Rural
|
Urban
|
Rural
|
Urban
|
|
Census – 2001
|
44
|
84
|
54
|
106
|
NSS (1999-00)
|
15
|
35
|
5
|
12
|
Census – 2011
|
48
|
61
|
79
|
94
|
NSS (2011-12)
|
14
|
23
|
6
|
11
|
Note: Census number
refer to non-workers seeking or available for work; NSS numbers are for those
termed ‘unemployed’
Source: Census of India
and NSS ‘Employment and Unemployment Situation in India’ reports
The National
Statistical Commission noted many years back that NSS estimates of unemployment
were low because of the definitions and concepts
followed in its framework of employment and unemployment.[18]
What are the
definitions and concepts in the NSS framework that makes its unemployment
estimates low compared to the census?
The NSS uses the ‘major
time’ criteria to decide if non-workers are ‘unemployed’ or outside the labour
force.[19]
Persons spending a major portion of their time in study or domestic duties are
considered to be outside the labour force. They do not add to the count of the
‘unemployed’ even if they are seeking work.
In the census, on the
other hand, the test of ‘seeking work’ is applied to all non-workers without
any hard time criteria associated with it. Persons spending a major portion of
their time in study or household duties can still qualify the test and be
counted among the census unemployed.[20]
Students and unemployment
The above arguments, to
the extent that they involve the double counting of students among the
unemployed in the census, can be substantiated by examining age disaggregated
data on non-working men (Table 4). The discussion can be limited to men in the
age group 15-29 without compromising the analysis as this group accounts for 85%
of the census unemployed.
Table 4. Age disaggregated profile of non-working men seeking
work gathered from Census 2011 and NSS 2011-12
Census
|
NSS
|
|||||
Age group
|
Not working
|
Of which
|
‘Unemployed’
|
|||
Students
|
Seeking work
|
* Students seeking work
|
||||
Rural
|
15-19
|
659
|
544
|
128
|
72
|
30
|
20-24
|
294
|
190
|
117
|
51
|
46
|
|
25-29
|
107
|
26
|
50
|
0
|
21
|
|
Urban
|
15-19
|
768
|
651
|
140
|
76
|
33
|
20-24
|
430
|
295
|
161
|
68
|
70
|
|
25-29
|
164
|
47
|
77
|
0
|
45
|
Notes: Numbers are
ratios per 1000 population; * Lower bound on number of students seeking work
Source: Census of India[21] and NSS “Employment and Unemployment
Situation in India” report
The census count of
those seeking work far exceeds the NSS count of ‘unemployed’ for each age group
as expected. Census numbers indicate that the bulk of non-working men of age
15-24 are students. Some of them can be expected to figure in the census count
of those seeking work. A lower bound on the number of students seeking work has
been computed and presented in Table 4 using the detailed breakup of
non-workers by ‘main activity’ published as part of the census. [22]
The NSS count of male
‘unemployed’ agrees more closely with the census numbers after students seeking
work have been excluded from the latter. It needs to be re-emphasised that what
is listed in Table 4 is only a lower bound and exact numbers of students
seeking work, if available, will further improve the fit between census and NSS
data on unemployment among non-students. Even with the available data, it is
clear that the census unemployment is much higher than NSS unemployment in
males predominantly because it factors in students seeking work.
Women engaged in domestic duties and unemployment
Census figures indicate
that one must consider women between 15 and 39 to account for 80% of female
unemployment. Since disaggregated data is not available for this range from
both census and NSS, the discussion below is with aggregated numbers. Table 5
presents data from the census and NSS for non-working women aged 15 and above.
NSS data shows that these women are overwhelmingly students or engaged in
domestic duties according to the ‘major time’ criteria.
Table 5. Profile of non-working women aged 15 and above
seeking work gathered from NSS 2011-12 and Census 2011
NSS
|
Census
|
||||||
Not working
|
Of which
|
‘Unemployed’
|
Women in DD
available for work^
|
Seeking work
|
of which
|
||
Engaged in domestic duties(DD)
|
Students
|
‘Non-students’ seeking work *
|
|||||
Rural
|
648
|
499
|
85
|
6
|
33
|
79
|
54
|
Urban
|
806
|
611
|
122
|
11
|
41
|
94
|
59
|
Notes: Numbers are
ratios per 1000 population; ^ Estimate of the number of non-working women usually
engaged in domestic duties who are willing to take up regular full time work at
home;* Upper bound on the number of ‘non-students’ seeking work
Source: Census of India[23]
and NSS report “Participation of women in specified activities along with
domestic duties”[24]
The census data allows
a rough division of the total number of women ‘seeking work’ between students
and ‘non-students’.[25]
Table 5 lists the upper bound on the number of ‘non-students’ seeking work. From
the NSS perspective, these non-students would almost entirely be women engaged in
domestic duties.[26]
Again from the NSS
perspective, non-working women who spend ‘major time’ in domestic duties would
not qualify as ‘unemployed’ even if they were seeking work. However, the NSS
recognizes this situation as a form of underemployment and provides several
indicators of its extent. One such indicator – the number of non working women,
engaged principally in domestic duties, that are willing to take up regular
full time work if offered such work at home – is also shown in Table 5.
It can be seen that there is a fair agreement between the census numbers of ‘non-students’ seeking work and the aggregate of the NSS numbers of ‘unemployed’ and women primarily engaged in domestic duties who are available for work at home full time. Too much should not be read into the agreement as the NSS and census ‘tests’ of the availability of women engaged in domestic duties for work are not quite the same.
Nontheless, the
available evidence makes it clear that census unemployment in women is far
higher than NSS unemployment predominantly because it factors in students
seeking work as well as women, primarily engaged in domestic duties, seeking
work.
Concluding remarks
The comparison of NSS
and census data reveals that the differences in employment estimates have
narrowed over the last three censuses, the only exception being the estimates
of rural women in Census 2011. This exception merits further study especially
because the NSS numbers have not been on expected lines.
The NSS provides
several indicators of underemployment among workers; however none of these are
strictly comparable with census data as the coverage of workers is different in
the two exercises.
Census unemployment estimates are far higher than the NSS estimates because they also include students and persons primarily engaged in domestic duties who are seeking work. The desire for work in students and in women primarily engaged in domestic duties is in some sense a reflection of underemployment in these sections. The census unemployment estimate, in this sense, is an aggregate estimate of the unemployed and underemployed.
While the
underemployment among women engaged in domestic work is an ongoing part of NSS
employment studies, it has been left to the census, with its relatively crude
methodology, to uncover the extent of the latent Indian youth labour force.
References:
Census (2011), “Instruction Manual for Updating of Abridged House list and
Filling up of the Household Schedule”, Census
of India, (http://censusindia.gov.in)
Hirway, Indira (2012):
“Missing Labour Force, An
Explanation”, Economic and Political
Weekly, vol xlvii no 37, Sept 15, 2012
NSSO (2008), “Review of Concepts and Measurement Techniques in Employment
and Unemployment Surveys of NSSO”, National
Sample Survey Organisation, Dec 2008, (http://mospi.nic.in)
NSSO (2014a), “Employment
and Unemployment Situation in India, NSS 68th round”, National
Sample Survey Organisation, Jan 2014, (http://mospi.nic.in)
NSSO (2014b), “Participation of Women in Specified Activities along with
Domestic Duties”, NSS 68th round”, National Sample Survey
Organisation, Sept 2014, (http://mospi.nic.in)
Thorat Amit (2004),
“NSS Employment Surveys; Problems with comparisons over time”, October 2004,
Viewed on 10 Feb 2015 (http://www.macroscan.org/anl/oct04/pdf/NSS_Data_Critique.pdf)
[1]
Census 2011 tables containing data on workers were released in Sept 2014. The
Census estimate of unemployment in urban men aged 15 and above exceeds the NSS
estimate by a factor of 2.6. The factors for rural men and rural and urban
women are much higher.
[2] Data
from Table B1, Census 2011
[3] The ‘reference
period’ is the period to which the data pertains
[4] For
a brief historical overview of the problems of comparing NSS and census
employment data starting from the 1951 census, see Thorat (2004)
[5] See
Sec 1.5 NSSO (2014a) for a description of survey sampling in rural and urban
areas
[6]
Census B series tables for various years are available at the Census of India
website (http://www.censusindia.gov.in)
[7] The
census figures of ‘seeking or available for work’ are popularly interpreted as
unemployment figures though it must be pointed out that the census itself
refrains from using the term ‘unemployed’.
[8] A
complete explanation of the ‘major time’ criteria may be found in Sec 2.21 of
NSSO (2014a)
[9]
These reports are available at the Ministry of Statistics and Program
Implementation website (http://mospi.nic.in/)
[10]
Ratios computed by author using population figures from the Table B1 for
various years
[11] Data
is from the ‘usual activity status (ps+ss)’ table
[12] Sec
10.1, NSSO (2008)
[13] Sec
9.4.34 (xiv) , NSC(2001)
[14] See Metadata of
Census 2001 at Census of India Website
[15] The
census and NSS have different minimum time criteria for categorising persons as
workers (30 days in NSS, none in census). Census data for 2011 shows that only
37 in 1000 rural women aged 15 and above were ‘marginal workers’ working less
than 3 months. Even if all of these were missed out by the NSS, it would account
for only a small part of the NSS underestimation.
[16] See Hirway (2012) for a summary of the various explanations attempted
[17] NSS data for 2011 is
from Table 61 of NSSO (2014a). The table provides estimates of ‘usual
status workers (ps+ss)’ who were without work for at least 1 month and who
sought or were available for work on at least some days during those month(s).
This number (N) is regardless of the age of the workers, and may include some
workers below 15 years of age. ‘N’ serves as an upper bound of this indicator
of underemployment among workers of age 15 and above. A lower bound is obtained
by subtracting the total number of ‘usual status (ps+ss)’ workers aged below 15
from N
[18] Sec
9.4.27 (a) of NSC(2001)
[20] Sec
6.166, Census(2011)
[21] Raw
data is from Tables B1 and B13 of Census 2011. Authors computation of the number of students seeking work is
explained in note 22
[22]
Table B1of the census provides the total number of non-workers seeking work
(U). Table B13 divides non-workers into categories by ‘main activity’ including
the categories of ‘students’ (S) and ‘others’ (O). According to the census
enumeration guidelines all ‘non-workers’ seeking work, other than students, are
placed in the category ‘others’. The ‘others’ category also includes some
persons who are not seeking work. (Sec 6.160, Census(2011)) ‘O’ is therefore an
upper bound on the number of non-students seeking work and (U-O) represents a
lower bound on the number of students seeking work
[23] ibid
[24] Table
6 of NSSO (2014b) provides the fraction of women aged 15 and above, usually
(principal status) engaged in domestic duties and without subsidiary status
work, who are ready to accept work at home. Table 5 provides the fraction of
those willing to work at home (irrespective of subsidiary status) that are
ready to accept regular full time work. The estimate of ‘women in domestic
duties available for work’ is arrived at by applying these two fractions to the
estimate of women usually engaged in domestic duties and without subsidiary
status work (Table 2)
[26] As
an example, after removing students from the rural ‘non-working’ women aged 15
and above, women engaged in domestic duties constitute nearly 90% of the rest
(Table 2, NSSO(2014b))
Links:
Another paper analyzing the same data sets as above - India's rural transformation: A myth or reality?
Links to this article:
1. Changes_in_indias_rural_labour_market_in_the_2000s
2. Positives_outweighing_negatives_the_experiences_of_Indian_crowdsourced_workers
3. Emerging_Trends_and_Patterns_of_India%27s_Agricultural_Workforce_Evidence_from_the_Census
4. indias-unemployment-rate-conundrum
5. Changing Employment and Enterprise Structure in Gujarat: 1990-2005
Links:
Another paper analyzing the same data sets as above - India's rural transformation: A myth or reality?
Links to this article:
1. Changes_in_indias_rural_labour_market_in_the_2000s
2. Positives_outweighing_negatives_the_experiences_of_Indian_crowdsourced_workers
3. Emerging_Trends_and_Patterns_of_India%27s_Agricultural_Workforce_Evidence_from_the_Census
4. indias-unemployment-rate-conundrum
5. Changing Employment and Enterprise Structure in Gujarat: 1990-2005
No comments:
Post a Comment