This paper seeks to examine the efficacy of the time criterion of unemployment in poor agrarian economies
from the point of view of identifying poverty groups. It indicates that even with the alterations and dis-
aggregations which have been recently suggested in the literature, the time measure of unemployment leaves
large gaps. Further, the unemployed persons need not necessarily be poor, irrespective of the category of
workers to which they belong. It is suggested that alternative criteria for measurement of unemployment, like
the productivity criterion, are likely to be useful supplements to the time based measurements, for getting a
holistic picture of unemployment and poverty.
Dr. Sarthi Acharya is Head of the Unit for Rural Studies, Tata Institute of Social Sciences, Deonar,
Bombay 400 088.
The extent of rural poverty in developing countries, in most of Asia, is much more
widespread than the recorded levels of rural unemployment. This had led to a
scepticism regarding the validity of the conventional time-criterion' of measuring
employment in agrarian settings from the point of view of economic utility of the work
done (Acharya, 1983; Raj Krishna, 1976; Sen, 1975). The fundamental relationship
that links work with wage (income) could be highly distorted in agrarian economies
since assumptions such as work-rationalization and non-zero marginal productivity,
need not hold. In view of this observation, a number of alternative measures have
been evolved to estimate surplus labour (Ahuja, 1975; Mehra, 1976). In a recent
paper, Visaria (1981) has suggested that, with some adaptation, the time criterion
approach itself can provide better estimates of the under-utilisation of the labour time
available. Visaria has demonstrated an association between unemployment and
poverty and has further shown that the bulk of the unemployed are concentrated
among the landless and the near landless labourers1.
This paper attempts to re-examine the adequacy of the time criterion of employment/
unemployment, to highlight the fact that, in spite of the adaptations in the data
classification, the time criterion measure is inadequate for explaining the extent of
poverty and for identifying approaches for poverty eradication2. More specifically, it
seeks to (a) examine the linkage between the extent of employment and levels of
living, (b) study the composition of the labour force to adjudge the employment and
levels of living in its different components, and, (c) decompose the unemployment by
components and levels of living to identify which categories of the unemployed are
poor. In no way is it, however, being suggested that the large surveys conducted to
measure idle time do not serve any purpose. The paper only intends to highlight that
the prevalence of the extent of underemployment a la Nurkse cannot be fully
captured by the time criterion.
This paper forms a part of an unpublished report submitted to the Indian Council of Social Science Research.
I am thankful to an anonymous referee who offered several useful suggestions on the earlier report.

126 Sarthi Acharya
The data used in this paper are drawn from the 32nd round of the National Sample
Survey (NSS) pertaining to 1977-78 for All-India.
The NSS uses three methods simultaneously for measuring the activity status of a
person. The activity on which a person spends major part of his/her time during the
365 days preceding the date of the survey is considered the usual activity status (US)
of the person. The second method is the current weekly status (CWS) which initially
records a unique activity status with reference to the week preceding the date of
survey. A person is classified as (un) employed if he/she has (not) worked for even
one hour during the reference week (and is actively seeking or is available for work).
The last method is the current day status (CDS) in which the unit of classification is a
'half day'. A person is classified as having worked a full day if he/she has worked for
4 hours or more in a day. He/she is classified as having worked for half a day if he/she
has worked for one hour or more but less than 4 hours in the day, and is unemployed
for half a day if he/she is seeking or available for upto 4 hours for work on that day. If
a person is not engaged in any gainful activity even for one hour during the day and is
available for work for 4 hours or more, he/she is classified as unemployed for the
entire day.
Table 1 shows the extent of unemployment in rural India as measured by the three
methods described above.
Table 1
Usual Status
2.12 5.52
Current Weekly Status
3.57 4.13
Current Daily Status
7.12 9.18
Source: GOI 1981
Poverty-Unemployment Linkages
The first relationship sought to be established here is the one between income
(expenditure)3 and the extent of employment and unemployment in rural India. For
this purpose, employment as per the CDS, which is the closest to the flow concept of
labour time, is tabulated by expenditure class in Table 2. It is evident from the table
that the incidence of employment (unemployment) is higher in the higher (lower)
expenditure brackets. The simple correlation co-efficient between the mean per
capita expenditure and employment of persons (i.e. row 3 of table 2), is 0.86, which is
highly significant. The span of variation in unemployment is, however, small, ranging
between 4 and 15 per cent, while the expenditure classes, at least theoretically, range
between 0 and 00. Some explanation of the gap between employment and levels of
living can be attributable to the economic dependency ratio, in view of the fact that
the expenditure in Table 2 relates to 'per capita expenditure of the household'. It is
not the individual income. However, even after adjusting the data for this factor, a
large gap still remains4.

Table 2

128 Sarthi Acharya
Data on the distribution of population by expenditure classes shows that 49.6 per cent
of the persons are below the poverty line while the extent of unemployment does not
exceed even 20 per cent.5 At the aggregate level, therefore, it is evident that the
relationship between expenditure levels and employment is strong but limited.
Attempt is, therefore, made to disaggregate the data by its occupational composition
because of the heterogeneity of the nature of work and workers.
In Table 3, we present data on the incidence of unemployment, as per CDS,
according to the type of households6 and its major source of income to scrutinize
whether the composition of the work force provides any explanation for the gap
between the reported incidence of poverty and unemployment.
Table 3
Household Type
Percentage Incidence of
Composition Unemployment
of the Total
Rural Work-Force M F P
Self-Employed in Agriculture
51.72 2.83 2.26 2.68
Self-Employed in Non-Agriculture
10.73 5.19 6.37 5.48
Agricultural Labourers
26.41 14.69 18.40 16.36
Other Labourers
6.48 12.50 13.18 12.73
4.65 8.31 10.86 8.80
100.00 7.12 9.18 7.70
The table shows the highest incidence of unemployment among the agricultural
labourers. The marked similarity in the unemployment incidence among the
households of agricultural labourers, the unemployment in the lower expenditure
groups shown in Table 2, and the fact that about 56 per cent of the total unemployed
are agricultural labourers, may lead one to conclude that poverty is concentrated
among the agricultural labourers, who also manifest high unemployment rates.
But Table 4, which presents a classification of the employed persons by CDS,
disaggregated by the household type, calls for more caution before drawing
conclusions.7 There is a large percentage of the poor among the employed in all
categories of jobs. It is true that the poor are less among the self-employed and are
more among the agricultural labourers, but the problem of poverty is substantial
among cultivators, who till the uneconomical^ small plots of land, or artisans whose
tools permit extremely limited productivity.
To adjudge the extent of poverty of the working poor, vis-a-vis the unemployed, a
co-efficient of the extent of poverty is defined.8 Let P be the said poverty line and Mo
be the mode of the frequency distribution of persons over the expenditure classes.
The measure of the extent of poverty is defined as:

130 Sarthi Acharya
The denominator is the standard deviation which incorporates the spread effect. The
variables f, x and N in the formula are the frequencies, the mid-values of the
expenditure classes and the total population, respectively, as in Table 4. The formula
is similar to the Pearsonion skewness co-efficient with the signs reversed, except that
the 'mean' is replaced by P. The formula follows the logic that the purpose here is to
determine the distance of the modal frequency from a fixed poverty-line, moderated
by the spread effect, in a number of similar shaped distributions, with a view to guage
their relative distances from P. 'E' will be positive in less groups and the magnitude
will rise with the fall in poverty. The converse is also true, i.e. in groups reporting a
large incidence of poverty, the index will be small and may even become negative.
With a rise in poverty, the index will rise with a negative sign. A zero value indicates a
situation where the mode coincides with the poverty line. It has no other significance.
Theoretically, the index can range between - 0 0 to +00. The values are computed for
the six categories of employment in Table 4. It is also computed for the umemployed,
for a contrast.
Table 5 shows that the extent of poverty among the working regular employees in
agriculture and working labourers in non-agriculture is similar to that among the
unemployed persons. Further, the extent of poverty among the working agricultural
labourers is much more than that among the unemployed workers. Thus, to say that
the poorest are so because they are unemployed, is not necessarily true. Some
employed are very poor, while some unemployed probably wait for better
Table 5
Note: Nos. 1 to 6 refer to the same household type categories as in Table 4. No. 7 refers to the unemployed.
Compostion of the Unemployed
Following from the above, attempt is now made to disaggregate the unemployed
persons by their household type and poverty status. The breakdown of the
unemployed persons by type of household and expenditure patterns, is as such, not
available. We, therefore, calculate these figures using an indirect method.
For simplicity's sake, consider that there are two categories of workers and only one
sector.9 The total unemployed persons can be categorized into those belonging to
self-employed households and are poor (X11), those belonging to the self-employed
households and are not poor (X12), those belonging to labourers' households and are
poor (X21) and those belonging to labour households and are not poor (X22). The
following identities can be then established.

Poverty-Unemployment Relationship in Rural India 131
Here U1 and U2 are the unemployed persons belonging to the self-employed and the
labour households, and P1 and P2 are the poor and the non-poor unemployed
workers. U1, U2, P1 and P2 are known from Tables 3 and 4, respectively. The system is,
however, still under-determined since U1 + U2 = P1 + P2. An additional condition
is necessary for a solution of the system. Let us hypothesize a relationship,
X11 = X12, where is the ratio of the quantum of unemployment reported by the
poor and the non-poor among the self-employed households.
The value of is obtained indirectly under restricted conditions. First, the
relationship between the levels of living and the size of land-holdings is assumed to
be continuous and monotonia Under this assumption it is possible to obtain the land
size holding at the poverty line from a two-way cross tabulated data on the levels of
living and land holding of cultivators. Let this be L. Next, the data on currently
unemployed persons whose principal occupation otherwise is agriculture, are
tabulated by their land size holdings. Under the assumption that land holders are
all cultivators by the usual status, the figures for the number of the poor and non-poor
unemployed cultivators can be calculated by demarking these data by L. The
numerical values of for male and female workers are 4.75 and 6.64, respectively.10
The solution of system (1) based on these values of a is given in Table 6.
Note: Figures in parentheses are the percentages to the total unemployment.

132 Sarthi Acharya
It is evident from this table, that among the unemployed persons belonging to the
labour households, some 26 per cent are not poor. Among the unemployed persons,
belonging to self-employed households, some 18 per cent are poor.11 It follows that
poverty is prevalent among the employed and the unemployed, in all types of
households, and vice-versa. This result thus points out that, in addition to the
agricultural labourers, some cultivators and other own account workers also face the
brunt of poverty during their period of unemployment. Small and uneconomical land
size holding may neither provide employment on a continuing basis nor yield
The time criterion, it is demonstrated, is not able to capture all the dimensions of
poverty in agrarian structures. In this regard, the productivity/income criteria of
unemployment can be used as alternative methods to supplement the existing time
Productivity Criterion Estimates
The productivity (or the income) criterion yields meaningful estimates from the point
of view of identification of poverty groups and determination of surplus labour in rural
agrarian settings. These cannot, however, be calculated from the groups NSS data.
Some of the estimation methods are laid down in Islam et al (1982) and Acharya
(1983). Here we present estimates from some studies for illustrative purposes.
The methods followed by different authors for measuring unemployment by the
productivity criterion can broadly be classified into two categories. The first method
entails the calculation of the 'required' labour for performing a certain operation,
given the technology. This is then subtracted from the total utilized labour to obtain
the unemployed (surplus) labour. The second method assumes that economically
meaningful employment is one which yields a marginal productivity at least equal to
the prevailing wage rate. All those persons whose marginal productivity (some
authors have used variations of average labour productivity) is less than the average
wage, or a stipulated income, are considered unemployed. For the sake of
convenience, we will call these as method A and B, respectively. In Table 7, some
estimates are provided as examples.
Table 7

Poverty-Unemployment Relationship in Rural India 133
It is evident from Table 7 that the productivity criterion of unemployment is closer to
explaining the poverty status. The gap which still exists could be due to the method
followed, since in method A, it is possible that any rationalization may not yield
statisfactory labour productivity due to the prevailing technology and agro-climatics.
The economic dependency ratio may also cause variations. In some cases, the
poverty is lower than unemployment. This is quite possible when some unemployed
persons are able to share the earnings of the employed and, thereby, avoid falling
below the poverty line.
This paper examines the relationship between poverty and unemployment in rural
India with a view to assess the efficacy of the time criterion of unemployment for the
identification of poverty groups. The relationship between poverty and
unemployment is found to be strong but limited. It is also seen that large numbers
among the employed are poor in all categories of workers, though the agricultural
labourers are found to be poorer than the cultivators. Further, not all unemployed
persons are poor even in the category of the labourers. The time criterion of
measurement is thus of limited importance in the identification of poverty in agrarian
societies. Alternative measures like the productivity/income criterion are suggested
to be more meaningful.
1. The 'adaptations' referred to by Visaria are, (1) disaggregation of the labour force and treatment
of each of the segments separately, and (2) usage of multiple definitions of time use. We have
used both these adaptations in this analysis, with the difference that data on one of the
definitions of unemployment is not available to us. This however, does not change the character
of the arguments presented here. The data used by Visaria are from the National Sample Survey
(NSS), 27th round, 1972-73. The NSS collects nationwide large samples on different facets of the
socio-economic situation of the country on a continuing basis in its different rounds. The data are
highly representative and authentic. The data are published in grouped forms in one way and two
way classifications on a variety of economic, social and demographic variables. Re-grouping or
cross tabulating is not always possible from the published tables. The raw data are kept
2. Visaria too admits that poverty is higher than unemployment, as a passing reference.
3. The NSS collects data on the household expenditure. Data on income are not collected. Expenditure is
used as a proxy for income.
4. The average dependency ratio across expenditure classes i.e. the proportion of economically active
persons in the household, is 0.41, with a standard deviation of 0.02. The deviation is so small that one
would easily take the dependency ratio to be 0.41 for the population. An adjustment with this ratio will
provide a crude estimate of earnings per worker if expenditure is equated to earnings. Even with
this adjustment, the argument does not change much. Moreover, every society has a dependent
population. Normatively, the quantity of earnings should be such that there is scope of feeding a 'normal
size' dependent population with the household earnings.
5. Poverty is defined as per the accepted norms laid down by the Government of India. (See GOI 1980).
6. The details of occupation/industry/status are standardized by the ILO, and similar definitions are followed
in most countries. Not all categories are applicable in all countries. Table 3 is the most typical
disaggregation of the composition of rural work force applicable in Indian conditions.

134 Sarthi Acharya
7. Table 4 has a minor modification in the disaggregation of the composition of the work force as compared
to Table 3. A category of 'employees' is introduced, which constitutes less than 10 per cent of the labour
force. About 4 per cent are in agriculture and 5 per cent in non-agriculture. The latter enjoy a better living
standard compared to any other category.
8. We are aware of the more sophisticated measures of the extent of poverty. The Sen index is a typical
example (see Sen 1980, Kakwani 1981). However, due to non-availability of data on the actual mean
values of the interval classes, the Gini co-efficient, which is a pre-requisite for computation of any of
these formulae, cannot be accurately computed.
9. This assumption is not very unrealistic keeping in view, the frequency distribution of the persons over the
expenditure classes in Table 4. Further, the size of the non-agricultural sector is small.
10. This method, at best, provides crude approximation of X since the levels of living and land holding need
not be continuous or follow a monotonic relationship. Next, it is not necessary that all land owning
persons are self-employed by usual status. In spite of these limitations the estimates can be acceptable
for such an aggregative exercise.
11. These are calculated from the row percentages not directly readable from the table.
Acharya, S.
"On the Productivity Criterion of Unemployment in Poor Agrarian
Economies" Development and Change, Vol. 14, No. 1.
Ahluwalia, M. S.
"Rural Poverty and Agricultural Performance in India", Journal of
Development Studies Vol. 14, No. 3.
Ahuja, K.
"Measurement of Rural Labour Surplus" Ph.D. Thesis, University of
Rajasthan, Jaipur.
Dandekar, V. M.
"On Measurement of Poverty", Economic and Political Weekly,
February 6.
GOI (Government of India)
Sarvekshna Journal of the NSS, New Delhi, Oct-Dec.
GOI (Government of India)
"Sixth Five Year Plan", New Delhi.
Houser, P. M.
"Measurement of Labour Utilisation" (Mimeo).
Islam, R.,
Employment and Development in Nepal, ILO-ARTEP Publication,
Khan, A. R. and Lee E.
Kakwani, N. C.
Income Inequality and Poverty, World Bank Publication, Washington.
Mehra, S.
"Measurement of Surplus Labour in Indian Agriculture", Indian Economic
Review, April.
Raj Krishna
"Rural Unemployment and Survey of Concepts Estimates for India" World
Bank Staff Working Paper, No. 234, Washington.
Sen, A. K.
Employment, Technology and Development, Oxford University Press,
New Delhi.

Poverty-Unemployment Relationship in Rural India 135
Sen, A. K.
Poverty and Famines, Oxford University Press, New Delhi.
Visaria, P.
"Poverty and Unemployment in India—An Analysis of Recent Evidence",
World Development, Vol. 9, No. 3.