Cambridge Journal of Economics 2016, 40, 843–870

doi:10.1093/cje/bev025
Advance Access publication 28 May 2015

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Jobless growth in India: an investigation
Sheba Tejani*

This paper investigates the relationship between output growth and employment
growth in India for the period 1978–2010 at the aggregate and sectoral levels.
Using a Kaldorian framework of endogenous productivity growth, we find that
Kaldor–Verdoorn effects in the economy have become more predominant over
time, especially in the post-reform (1994–2010) period. Our estimated Kaldor–
Verdoorn coefficients, measured as the employment elasticity of output growth,
for both formal sector and total employment have dropped dramatically over time,
suggesting that India has leapfrogged into a high-productivity regime without the
broad-based expansion of labour-intensive production that has been characteristic
of fast-growing economies in East Asia. We examine some explanations for why
these Kaldor–Verdoorn effects have become pronounced over time and are not con-
vinced that wage pressure has been one of the reasons. A shift in the composition of
demand towards higher-productivity sectors, however, does appear to be an impor-
tant part of the explanation. We also find mixed evidence that forces of international
competition have generated pressures to adopt more capital-intensive techniques of
production.

Key words: India, Jobless growth, Kaldor, Composition of demand, Employment

growth
JEL classifications: J20, O10, O40

1. Introduction
The Indian economy has grown at an average rate of over 7% since the year 2000 and
remains one of the fastest-growing countries in the world, although there has been a
slowdown recently. India also displays a skewed output–employment structure: services
make up the largest share of output while employment is concentrated in agriculture
and mostly informal in nature.1 Thus although a kind of structural transformation has
been achieved in that resources have been reallocated to more productive sectors, the
transfer of labour to these sectors as predicted by theories of development is far from

Manuscript received 7 September 2013; final version received 14 February 2015.

Address for correspondence: 1 Parade Ground Road, Camp, Pune 411001, India; email: shebatejani@gmail.com

*  Graduate Program in International Affairs, The New School, New York. I would like to thank Prof.
Duncan Foley for his advice and support through the process of writing this paper. Thanks are also to Prof.
William Milberg and Prof. Sanjay Reddy for providing valuable comments and suggestions. Any errors and
omissions remain mine alone.
1
  Informal employment is approximately 92% of the total (NCEUS, 2009). NCEUS makes a distinction
between the ‘informal sector’ and ‘informal workers’. Informal workers include those employed in the infor-
mal sector and those workers in the formal sector who do not receive any employment and social security
benefits.
© The Author 2015. Published by Oxford University Press on behalf of the Cambridge Political Economy Society.
All rights reserved.
844  S. Tejani
being realised. This structure has obvious implications for inequality: the gap between
agricultural and non-agricultural incomes has widened over time as the majority of

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
the labour force is concentrated in shrinking and low-growth agriculture.2 Although
economic growth has delivered prosperity for some, a large number of farmers in rural
areas have committed suicide due to indebtedness and crop failure; rural infrastructure
remains inadequate and rising inflation has eaten into the purchasing power of workers.
Consumption inequality has been on the rise in urban areas since the 1980s and the
gap between urban and rural mean consumption levels has been growing over the long
term (World Bank, 2011, pp. 11–12). Although India’s Gini coefficient of inequality is
considered low by international standards, when income rather than consumption data
are used to estimate inequality, its level of inequality is comparable to that of Brazil and
Argentina and higher than China’s (World Bank, 2011, pp. 23–4). In this situation, the
lack of viable employment opportunities is not only a driver of inequality, but also rep-
resents a growing social and political problem in the country.
The previous government belatedly recognised this problem by enacting a rural
employment guarantee scheme (Mahatma Gandhi National Rural Employment
Guarantee Act, MGNREGA) in 2005 that guarantees 100  days of employment to
each rural household per year. A  limited employment guarantee act can serve as a
much-needed temporary reprieve from unemployment, but it can hardly be expected
to address the larger structural problems the country faces, namely a persistent imbal-
ance in the output–employment structure, a low-growth agricultural sector and a
largely informal labour market.
The question is whether the recent spell of economic growth is exacerbating or ame-
liorating these structural imbalances and, further, whether India has leapfrogged into a
high-productivity regime that is not conducive to employment growth. If this is the case,
supply-side policies such as building human capital and providing vocational training and
job-matching services3 will clearly be insufficient to deal with the shortfall of employment in
the non-agricultural sectors and indeed even to stimulate growth in the agricultural sector.
Theoretically, productivity growth has been considered the sine qua non for both
continuing economic growth and employment generation. Sundaram and Tendulkar
describe the predicted linkages as follows:
Sustained expansion of productive activity that constitutes economic growth generates gainful
employment opportunities with continuously rising productivity. This makes possible a progres-
sive absorption and integration of the working poor into expanding economic activities often
involving rising productivity in their existing occupation with better technology or shift to new
occupations with upgraded skills. (Sundaram and Tendulkar, 2002, p. 6)

However, if productivity growth outstrips demand growth, employment growth will

decline. Productivity growth then has the perverse effect of displacing more labour
than it is absorbing and the question is what happens to the workers that are displaced
in the process. In countries with large intersectoral productivity gaps, it is likely that
workers displaced from relatively high-productivity sectors end up in low-productivity
and/or informal sectors (McMillan and Rodrik, 2011).4
2
  For example, in 1960 the ratio of agricultural to non-agricultural income was 1:3 which widened to 1:6
in 2005; the ratio of agricultural to industrial income was 1:4 while that of agricultural to services income
was 1:7 in per capita terms (Papola, 2007).
3
  See, e.g., World Bank (2007, p. 36).
4
  McMillan and Rodrik (2011, pp. 65, 67) find evidence for the same in Latin America and sub-Saharan
Africa for the period 1990–2005 and suggest it may be the result of a process of ‘industry rationalization’
spurred by trade liberalization.
 Jobless growth in India   845
There has been some disagreement in the literature and in policy circles on
whether Indian growth has been ‘jobless’5 and, more broadly, on whether economic

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
reforms undertaken in the country have been inclusive.6 There are differing views on
the period of jobless growth, on whether this period has ended, on its causes and on
predictions for the future.7 Most of the studies focus on the manufacturing sector or
the formal sector for which there are continuous time series data though, considering
its relatively small share in total output and employment, more cross-sectoral studies
are needed.
In this paper we find that the present scheme of Indian growth is unlikely to lead
to a rapid transformation of the employment pattern and that the underlying forces
responsible for this are getting worse, not better. Using a Kaldorian framework of
endogenous demand-led growth, we assess the relationship between output growth
and employment growth (formal and total) in India at the aggregate and sectoral lev-
els. Endogenous productivity changes are driving a slowdown in employment and they
appear as decreasing Kaldor–Verdoorn (K-V) coefficients8 in the economy, suggesting
that growth has become jobless over time. We find that a shift in the composition of
demand, rather than wage pressure or rising international competition, explains the
decline in K-V coefficients. Depending on data availability, the period under consid-
eration is divided into two subperiods, pre-reform and post-reform, in order to disen-
tangle the impact of neoliberal reforms that were undertaken in India after it suffered
a balance-of-payments crisis in 1991.

2.  Okun’s law and the Kaldor–Verdoorn effect

There are two distinct theoretical traditions that relate output growth to employment
growth. Within the orthodox tradition, Okun’s law relates quarterly changes in the
growth rate of GDP to the rate of unemployment, following Okun’s (1962) paper
that established a negative and statistically significant relationship between the two
variables. Okun tested two functional forms of the relationship: a difference model, in
which the quarterly percentage change in the unemployment rate was regressed on the
quarterly percentage change in real GDP, and a gap model, in which the deviation of
the level of output from potential output was related to deviation of the unemployment
rate from its full employment level. Okun’s law remains the dominant framework to
forecast the impact of output growth on employment within standard macroeconom-
ics and subsequent studies use a dynamic version of Okun’s difference formulation
that includes lagged values of both the independent and dependent variables (Knotek,
2007; IMF, 2010; Basu and Foley, 2013).
The heterodox tradition is based on Kaldor’s interpretation of Verdoorn’s (1949)
finding of a positive and statistically significant relationship between the growth of
industrial output and labour productivity growth for a sample of advanced indus-
trial countries. Kaldor combined Young’s (1928) insights on increasing returns as
a ‘macrophenomenon’ with Verdoorn’s (1949) empirical findings and claimed that

5
  See, e.g., Dasgupta and Singh (2005) and Papola (2007).
6
 The recent debate between Jagdish Bhagwati and Amartya Sen on whether India’s growth model has
been successful is one example. See Bhattacharya (2013) for a quick recap.
7
 See Bhalotra (1998), Goldar (2000), Nagaraj (2000, 2004), Rani and Unni (2004), Bhattacharya and
Sakthivel (2004) and Kannan and Raveendran (2009).
8
  The coefficient that relates output growth to employment growth. See Section 2.
846  S. Tejani
output growth itself induced shifts in productivity and not the other way around.
Manufacturing activities in particular, he argued, experienced increasing returns as

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
the scale of production expanded. Drawing on Smithian ideas, Kaldor contended
that as the market expanded, the division of labour became more pronounced and
translated into a greater specialisation and differentiation of production. This was
associated with higher returns per unit labour or greater labour productivity because
of the phenomena of learning by doing and static and dynamic economies of scale
(Kaldor, 1966, p. 105). Kaldor’s interpretation of Verdoorn’s law or the ‘K-V effect’
is the core mechanism through which labour productivity is spurred endogenously
through output growth. This can lead to a virtuous cycle of economic growth in the
presence of an elastic labour supply.
Empirically, Kaldor (1967) specified this relationship as a regression of productivity
growth on output growth or, alternatively, of employment growth on output growth,
where pt is productivity growth, yt is output growth and et is employment growth at
time t:

pt = − α + β1 yt (1)

et = α + β2 yt (2)

where β1 + β2 = 1
Kaldor (1966, 1967) found that the coefficient relating output growth to productiv-
ity growth for advanced industrial countries over a 10-year period from 1953 to 1963
was positive and significantly less than unity (approximately 0.5). For the purposes of
this paper, we focus on equation (2) and term β2 our K-V coefficient.9 Pieper (2003)
tests the K-V relationship in the form of equation (2) and finds evidence of increasing
returns10 at the sectoral level for a sample of 30 countries for a period covering the
mid-1970s to the mid-1990s, confirming that K-V effects are by no means limited
to the manufacturing sector. We use this framework to analyse both formal and total
employment (including informal employment).
A Kaldorian framework is suitable for the present analysis as it directly examines
the relationship between output growth and employment growth, whereas Okun’s
law relates output growth to changes in the unemployment rate, which is mediated
by changes in the labour-force participation rate. Thus ostensibly, as Basu and Foley
(2013) point out, the Okun coefficient could change as a result of changes in the
labour-force participation rate although employment per se has not shifted. K-V effects
can be examined at a disaggregated level while Okun’s law is not readily amenable to
such analysis for it requires the use of labour-force participation and unemployment
rates that are not meaningful at the sectoral level. Finally, the Kaldorian framework
offers a sound theoretical basis for relating productivity growth to output growth with
its resulting implications for employment growth, a virtue that Okun’s law lacks (Basu
and Foley, 2013).

9
 Following Basu and Foley (2013), we term the coefficient that relates output growth to employment
(rather than productivity) growth the K-V coefficient as we are interested in the impact of output growth
on employment.
10
  The coefficients are positive and less than unity.
 Jobless growth in India   847
3.  Formal employment and output
The question of whether the economy is generating formal sector jobs is important

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
because they are generally associated with higher pay, greater job security, superior
working conditions and better benefits. From a theoretical perspective, both neoclas-
sical and heterodox, the shift to a higher-productivity regime is expected to generate
higher-quality jobs.
We use data on ‘organised’11 employment across nine sectors at the one-digit level,
collected by the Employment Market Information Programme of the Directorate
General of Employment and Training (DGET), Ministry of Labour, India (MoF,
2011). The anomalous character of output and employment growth in India becomes
obvious in Figure 1: from 1991 to 2008, the trend growth rate of output was a healthy
6.2% while that for formal employment was −0.2%. In a country that is home to the
world’s second largest labour force, not only is formal employment (at 27 million)
relatively miniscule, it is still comprised largely of public sector employment (64% of
total) because private sector employment growth has been a non-starter. Since the
wave of deregulation and privatisation in the 1990s, the decline of public employment,
at an average of −0.72% over the period, has barely been compensated by the average
increase in employment in the private sector at 0.81%. One of the defining character-
istics of public sector employment in India is its overwhelming male bias—over 80%

Fig. 1.  Employment and growth, India aggregate, 1991–2008.

Sources: Author’s illustration based on CSO (2011a, 2011b) and MoF (2011).

11
  Official statistics in India use the term ‘organised’ employment instead of the more widely used ‘for-
mal’ employment. For consistency, the term ‘formal employment’ is used throughout this paper. Organised
employment is defined as employment in all public sector establishments and non-agricultural establish-
ments in the private sector with 10 workers or more. It does not include employment in agricultural and
allied activities, except in plantations. See http://www.dget.nic.in/content/students/employment-market-
information-program.php for coverage details.
848  S. Tejani
of employees in 2008 were men—with the result that women remain relegated to rela-
tively poorly paid informal sector jobs. This is less true for the private sector, though

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
women’s employment rates remain generally low.
We estimate the K-V coefficients by subtracting the trend growth rate of GDP from
the trend growth rate of formal employment for each sector as annual growth rates
tend to be noisier. The derivation is presented below, where E is employment and Y is
output:
ΔE = (ΔE/ΔY) * ΔY
If η = ΔE/ΔY or the K-V coefficient
ΔE = η * ΔY
Subtract ΔY from both sides:
ΔE− ΔY = η * ΔY − ΔY
ΔE− ΔY = ΔY(η − 1)
We estimate ΔE – ΔY and derive η:
ΔE− ΔY ≈ (ΔE/ΔY) − 1 = η− 1
Thus η = (ΔE – ΔY) + 1

First, as we see in Table 1, a number of sectors have shed labour even as they have grown
rapidly, namely utilities, construction, manufacturing and transport, storage and com-
munications. This decline was mostly concentrated in the public sector, though the
decline in employment in manufacturing is particularly striking in this regard because
it is mostly a private sector employer and has been growing fast. The FIRE sector expe-
rienced the highest growth of employment in the private sector at 7.7%.
We present our estimated K-V coefficients for the aggregate economy and by sector
in Table 2. The coefficients are estimated for the period 1991–2008 and by subperiods
1991–2000 and 2000–08 in order to detect shifts in the coefficient over the two decades.
Two salient points stand out: the decline in the coefficients over time is the result of a
cutback in the public sector and this decline speeded up in the second decade. Mining
and the financial intermediation and real-estate (FIRE) industries are the only sectors in
which output has been relatively employment intensive—their K-V coefficients over the
entire period are approximately 0.6 and 0.5, respectively—though the impact of mining
is limited in the aggregate as its share in total output and employment is relatively small.
They are also the only sectors that display non-negative growth in public employment
over the period, as evident in Table 2. Although employment contraction is not as perva-
sive in the private sector, and some service sectors, particularly the FIRE industries, have
contributed to positive employment growth, the magnitude of the increase has not been
enough to offset the massive decline in public employment.
Technical change considerably speeded up in the second decade under consideration,
so that in the construction and transport, storage and communications sectors where
output growth was very high and formal employment contracted, labour productiv-
ity growth was greater than unity. This implies that the K-V coefficients were actually
negative or that employment contracted when output expanded. In fact, employment
growth in four out of nine sectors from 2000 to 2008 was negative although all sectors
were growing at over 5%. Not only was output growth in agriculture the slowest, the
transfer of labour from agriculture to other sectors that was occurring in the 1990s
reversed itself in the later decade. In the aggregate, the K-V coefficient dropped dra-
matically from 0.47 in the 1990s to 0.2 in the 2000s, driven by negative employment
growth in a number sectors, on account of the contraction of public sector employ-
ment and stagnation in private sector employment.
 Jobless growth in India   849
Table 1.  Trend growth rates of private, public and total formal employment, 1991–2008

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Sector Private Public Total employment GDP
employment employment

Agriculture, forestry and 0.006 −0.011 0.000 0.027

fishing
(0.002) (0.001) (0.002) (0.001)
Mining and quarrying −0.008 0.007 0.006 0.043
(0.010) (0.005) (0.005) (0.001)
Manufacturing −0.002 −0.040 −0.011 0.066
(0.003) (0.002) (0.003) (0.002)
Utilities 0.014 −0.009 −0.007 0.059
(0.004) (0.002) (0.002) (0.002)
Construction −0.006 −0.022 −0.021 0.073
(0.010) (0.002) (0.002) (0.004)
Trade, hotels and 0.011 0.009 0.010 0.082
restaurants
(0.005) (0.002) (0.004) (0.001)
Transport, storage and 0.039 −0.011 −0.010 0.089
communications
(0.002) (0.002) (0.002) (0.003)
FIRE 0.077 0.008 0.027 0.079
(0.010) (0.002) (0.004) (0.002)
Community, social and 0.017 −0.003 0.000 0.062
personal services
(0.002) (0.002) (0.001) (0.001)
Total 0.008 −0.007 −0.002 0.062
(0.002) (0.001) (0.001) (0.001)

Table 2.  K-V coefficients for formal employment, 1991–2008

K-V coefficient

Sector 1991–2000 2000–2008 1991–2008

Agriculture, forestry and fishing 0.669 0.763 0.732

Mining and quarrying 0.451 0.938 0.629
Manufacturing 0.415 0.037 0.235
Electricity, gas and water supply 0.386 0.181 0.339
Construction 0.434 −0.344 0.061
Trade, hotels and restaurants 0.298 0.150 0.287
Transport, storage and communications 0.297 −0.351 0.018
FIRE 0.382 0.576 0.478
Community, social and personal services 0.432 0.363 0.379
Total 0.473 0.209 0.356

The question of why formal employment in the private sector has not increased has
been debated in the literature. Some of the reasons have to do with greater interna-
tional trends towards the flexibility of employment and the fragmentation and out-
sourcing of both manufacturing and services production (NCEUS, 2009). In the
Indian context in particular, a common argument is that labour market rigidity has
led firms to adopt intensive techniques of production and flexible labour practices
850  S. Tejani
(see Besley and Burgess, 2004; Fallon and Lucas, 1991), though this claim is vigor-
ously contested by others (see Bhattacharjea, 2006, 2009; Bhalotra, 1998). In par-

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
ticular, Kannan and Raveendran (2009) claim that factor substitution cannot be the
reason for the slow growth of formal employment as the cost of capital relative to
labour shows a rising trend over time while the share of wages in value added has
declined. The bias of both domestic and export demand towards goods that cater to
higher-income groups and have higher capital and skill intensity is another argument
(NCEUS, 2009; Patnaik, 2009). We return to some of these questions in Section 5,
where we attempt to explain our finding that output growth has been generating slower
employment growth over time.
Our findings here raise several issues. First, it could be argued that although formal
sector K-V coefficients have fallen over the period, informal employment could have
risen and even compensated for this decline so that growth would not have been job-
less. Second, formal employment data from the DGET are known to have some draw-
backs in that they reflect voluntary reporting by firms and have limited coverage. In
effect, it could be argued that the decline in the coefficients we are seeing is a statistical
artefact and may not reflect trends in the economy as a whole. Third, the distribution
of employment across urban and rural areas and by gender warrants some investiga-
tion. Hence we supplement our analysis by examining total employment data over a
longer period in the next section.

4.  Total employment and output

4.1  Sectoral and aggregate shares of employment and output
We begin by presenting some descriptive statistics based on the National Sample
Survey Office’s (NSSO) quinquennial (thick) employment surveys from 1978 to 2010.
The NSSO’s definition of an employed person includes those who are self-employed,
regular wage or salaried workers and casual labourers (NSSO, 2010, p. 10). Separate
employment series are created by classifying the employed according to their activities
during different survey reference periods. We use data on ‘usual principal status’ (UPS)
employment, which records the number of persons employed according to the main
activity pursued during a reference period of 365 days prior to the survey.12 Unlike in
many other countries, the NSSO’s measure of employment includes a large compo-
nent of informal employment for both principal and subsidiary status work and thus
estimates produced using these measures cannot be readily compared across countries.
Principal status employment is also more of a stock rather than a flow measure of
employment, in contrast to measures in other countries.13 The inclusion of informal
employment in the measure of total employment however is quite relevant from the
point of view of K-V effects, as Kaldor emphasised:
It is my contention that it is the rate at which this transfer [of labour from low- to high-pro-
ductivity sectors] takes place which determines the growth rate of productivity of the economy
12
  The other activity status categories include usual subsidiary status, which records the secondary activity
pursued by an employed person for not less than 30 days during a reference period of 365 days prior to the
survey. Current weekly status and current daily status records a person’s activity for a reference period of
one week and one day before the survey respectively (NSSO, 2010, pp. 12–14). Estimates of unemployment
are generally higher using current weekly and daily status.
13
  I am grateful to an anonymous referee for pointing this out. ‘Current daily status’ estimates of employ-
ment would be considered a flow measure instead.
 Jobless growth in India   851
as a whole. The mechanism by which this happens is only to a minor extent dependent on the
absolute differences in the levels of output per head between the labour-absorbing sectors and the

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
surplus-labour sectors. The major part of the mechanism consists of the fact that the growth of
productivity is accelerated as a results of the transfer at both ends—both at the gaining-end and
the losing-end; in the first, because of increasing returns, productivity in industry will increase
faster, the faster output expands; in the second because when the surplus-sectors lose labour, the
productivity of the remainder of the working population is bound to rise. (Kaldor, 1968, p. 386)

Informal employment is not only located in ‘surplus-labour’ sectors in India but is

also an integral part of formal sector production and thus needs to be captured in
order to track the spread of K-V effects through the economy. For example, the entire
increase in employment in the formal sector between the years 2000 and 2005 was
of the informal kind (NCEUS, 2009). Further, the informal sector itself can experi-
ence labour productivity growth not just through the loss of surplus labour as Kaldor
described, but also through the expansion of output and other enabling factors such as
trade liberalisation (Nataraj, 2011), greater access to technical knowhow and a rising
capital intensity of informal production (NCEUS, 2009). Complementary or com-
petitive relations between the formal and informal sector can also impact aggregate
labour productivity growth. Recent research has shown that formal manufacturing
firms in India have greater forward production linkages with ‘modern informal firms’
than with traditional ones and have in fact promoted their greater modernisation
(Moreno-Monroy et  al., 2012). Formal sector manufacturing firms have also ben-
efitted in terms of growth, productivity and competitiveness due to the lower labour
costs obtained through subcontracting, especially in the context of global value chains
(Unni and Bali, 2002). On the other hand, modern informal firms may become direct
competition for smaller formal sector firms when they grow in size due to their lower
operating costs. Examining the sources of labour productivity growth in the informal
sector as well as the nature of the linkages between the informal and formal sectors
are beyond the purview of this paper, for which disaggregated enterprise survey data
and case study methods are more suitable, though this is indicated as an area for
future research.14 Nevertheless, it is important to bear the complex interrelationships
between the formal and informal sectors in mind as we study K-V effects.
Our data are classified by industry using the National Industrial Classification (NIC)
for different years and presented for eight sectors; because changes in classification are
negligible at the one-digit level, data remain comparable across years.15 The NSSO
presents employment data as distributions per 1000 persons; these numbers are com-
bined with population estimates from the United Nations (UNSD, 2011) to obtain
the final estimates. Since the NSSO’s surveys are typically conducted mid-year (e.g.
July 2009 to June 2010), quarterly GDP estimates at factor cost and constant 2004–
05 prices (CSO, 2011a, 2011b) are used to construct output estimates for the same
14
 There are also some difficulties with the estimation of informal sector GDP. NCEUS (2009, pp. 1–2)
points out, ‘there has not been any enterprise survey in the unorganized sector for the estimation of gross
value added per worker in the agricultural, mining and electricity, gas and water supply sectors during the
last two decades. In the case of other sectors, the estimates are not available for all the industry groups for
any single year, as enterprise surveys on different subjects are conducted on different years.’
15
  Data for the year 2009–10 are classified according to NIC (2004) while years 2004–05 and 1999–2000
use NIC (1998). The year 1993–94 is classified according to NIC (1987) while years 1987–88 and 1983
use NIC (1970). NSSO (2011, p. 88) notes: ‘It may be noted that the changes in the activity classification
across [survey] rounds at one digit level of industry is nominal and thus comparison can fairly be made over
the rounds at the one digit level. This fact may be taken into account in measuring the changes of workforce
at one digit level over rounds.’
852  S. Tejani
period. The only exception to this is the year 1983, during which the NSSO survey ran
from January to December. The GDP estimates for this year are matched accordingly.

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Table 3 presents the share of employment and output by sector in 2010 sorted by
employment. ‘Other services’, which includes the FIRE industries and community,
social and personal services, is now the single largest contributor to GDP after having
overtaken the agricultural sector in 1998, though it contributes only 11% to employ-
ment. Despite agriculture’s nominal share of output, it remains the mainstay of the
workforce accounting for 49% of total employment; the two other largest employers,
manufacturing and trade, hotels and restaurants, comprise a much smaller share of
employment at about 12% each.
Another important feature of employment in India is its largely rural character,
which stood at 71% of total employment in 2010. After agriculture, construction is the
second largest employer in rural areas. In urban areas, other services, trade, hotels and
restaurants and manufacturing are the main employers and contribute approximately
25–23% each to total employment.
The female share of employment remains abysmally low and, what is worse, steadily
declined from 28 to 24% of total employment from 1978 to 2010. Female employ-
ment also remains a largely rural phenomenon and currently comprises about 79% of
total female employment, 79% of which is concentrated in agriculture. In summary,
the labour market in India remains highly segmented along urban and rural lines,
along agricultural and non-agricultural sectors and by gender. The workforce is largely
rural and male and sectorally concentrated in agriculture, followed at a distant second
by manufacturing and the trade sectors.

4.2  The Kaldor–Verdoorn effect

As employment data are available at five-yearly intervals (thick surveys) since 1978
and are expected to be less noisy than annual data, we use point estimates in order to
estimate K-V coefficients for total employment. We also assess the impact of economic
reforms by assessing shifts in the two subperiods: 1978–94 and 1994–2010.

Table 3.  Usual principal status (UPS) employment and GDP by sector, 2010

Sector UPS employment Share in total GDPa Share in total

(Rs. million)

Agriculture, forestry and 211,890,000 49 6,566,290 16

fishing
Trade, hotels and restaurants 51,261,824 12 6,967,870 17
Manufacturing 49,978,640 12 6,596,710 16
Other services 48,556,476 11 12,834,590 30
Construction 41,834,312 10 3,357,170 8
Transport, storage and 20,776,776 5 4,017,620 10
communications
Mining and quarrying 2,817,654 1 995,040 2
Utilities 1,252,993 0 843,270 2
Total 428,368,675 100 42,178,560 100

a
GDP in constant 2004–05 rupees.
 Jobless growth in India   853
4.2.1  Sectoral analysis of the Kaldor–Verdoorn effect 
Figure 2 displays the point estimates of the K-V coefficient at the aggregate level,

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
which have dropped sharply over time. From a high of 0.48 in 1983–84, the K-V
coefficient trended down to a low of 0.06 in 2009–10. Figure  3 presents point
estimates of the K-V coefficients by sector for the same period. Visual inspection
of the plots shows that all sectors other than construction, mining and utilities
display a declining trend in their K-V coefficients. In the agriculture, manufac-
turing, transport, storage and communications and trade, hotels and restaurants
sectors, coefficients trended down to zero or negative values by 2009–10. The
jump in the coefficients in the utilities and mining sectors in 2004–05 was caused
by sharply negative employment growth in the year 2000, which inflates employ-
ment growth in the next period due to a low base effect. In fact, employment in
utilities has not recuperated to its 1994 level despite rapid growth in output dur-
ing the period.
Employment in construction has shown large increases over time due to the rapid
urbanisation taking place in the country and particularly between 2004–05 and 2009–
10, due to the nationwide implementation of the MGNREGA that creates employ-
ment in public construction works.16 The other services sector does not show a clear
trend over time. It is noteworthy that in 2009–10, five sectors showed near-zero or
negative employment elasticities, which gives us an idea of how rapidly K-V effects
have spread through the economy.

Fig. 2.  K-V coefficients in the aggregate, 1983–2010.

Source: Author’s illustration based on CSO (2011a, 2011b) and NSSO (various
years).

16
  In 2009–10, 24% of all rural households obtained jobs in MGNREGA works, a significant number
(NSSO, 2011).
854  S. Tejani

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Fig. 3.  K-V coefficients by sector, 1983–2010.
Source: Author’s illustration based on CSO (2011a, 2011b) and NSSO (various
years).

Although the visual plots indicate a fall in the K-V coefficients over time, we fit a
linear time trend to the data in order to test if these trends are statistically significant.
Table 4 presents the trend growth rates and standard errors of sectoral and total K-V
coefficients for the entire period. The K-V coefficient for the aggregate has a negative
and statistically significant slope as does that for the transport, storage and communi-
cations sector; the agricultural sector is significant only at the 10% level. Though the
growth rates of the other sectors are not statistically significant, which may be partly
due to the averaging out of variability in the data in the aggregate, small sample esti-
mates tend to be unbiased and the slopes can serve as a reasonable first estimate of
the trends. Thus even using a broader measure of employment, it is evident that K-V
coefficients have fallen over time.

4.2.2 The Kaldor–Verdoorn effect by category of employment  In this subsection, we esti-

mate Kaldor’s structural equation for India by using the specification presented in
equation (3) and constructing a panel dataset by sector, as the number of years for
which we have observations is limited. We use a fixed-effects model to control for the
unobserved characteristics of each sector and are able to estimate the K-V coefficients
for the entire period and for the two subperiods under consideration:

eit = −α + β2 yit (3)

where i = sector
 Jobless growth in India   855
Table 4.  Trend growth rates of K-V coefficient by sector

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Sector b/se t/p R2

Agriculture, forestry and fishing −0.171 −2.35 0.642

(0.07) 0.08
Construction −0.767 −1.72 0.491
(0.45) 0.16
Utilities −0.067 −0.9 0.072
(0.07) 0.42
Manufacturing −0.079 −1.19 0.256
(0.07) 0.3
Mining and quarrying −0.067 −0.71 0.069
(0.09) 0.52
Other Services −0.053 −1.8 0.165
(0.03) 0.15
Trade, hotels and restaurants −0.077 −1.25 0.348
(0.06) 0.28
Transport, storage and communications −0.082* −4.21 0.73
(0.02) 0.01
Total −0.068* −2.93 0.659
(0.02) 0.04
Observations 6

b = beta, se = standard error, t = t-statistic, p = p-value. *p < 0.05, **p < 0.01, ***p < 0.001.

We estimate the coefficients for each category of employment, total employment, rural
employment, urban employment and female employment separately and present the
results in Table 5. The variables are transformed into logs so that the coefficients can be
interpreted as elasticities. Standard errors are robust to heteroscedasticity and intrasec-
toral correlation. The coefficients for all employment categories are highly significant
and less than unity, as Kaldor predicted, except for that on female employment in the
pre-reform period. The K-V coefficient for total employment has fallen dramatically
over the period from 0.74 in 1978–94 to 0.44 in 1994–2010, thus confirming that tech-
nological change speeded up considerably in the post-reform period. All other employ-
ment categories also show similar declines over the two periods, with urban employment
in particular displaying the lowest K-V coefficient in the second subperiod.

4.2.2.1  Identification issues.  Regressing output growth on employment growth,

on the other hand, yields coefficients that are all greater than unity, suggesting
that output growth regulates employment growth and not the other way around.
Nevertheless, there has been some debate on the possible endogeneity of output
growth in Kaldor’s model as the mechanism of cumulative causation suggests that
labour productivity growth feeds back into output growth in a virtuous cycle of
expansion (Rowthorn, 1975; Skott, 1999). If output growth is indeed endogenous,
ordinary least squares (OLS) estimates of the structural model will be inconsistent
and simultaneous equation methods or instrumental variables estimation will be
required.We explicitly test for the endogeneity of output growth in Kaldor’s structural
model by conducting an instrumental variables (IV) estimation. Following Basu and
856  S. Tejani
Table 5.  Fixed-effects regression analysis by category of employment, 1978–2010

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
1978–1994 1994–2010 1978–2010

b/se t/p b/se t/p b/se t/p

DV: total employment

GDP 0.735*** 7.68 0.439** 3.57 0.500** 5.27
(0.1) 0 (0.12) 0.01 (0.09) 0
Constant −3.907 −1.49 4.268 1.22 2.526 0.95
(2.62) 0.18 (3.5) 0.26 (2.65) 0.37
R2 (overall) 0.943 0.776 0.764
Endogeneity 0.273 0.227 0.224
DV: rural employment
GDP 0.853*** 7.4 0.450* 2.71 0.513** 4.47
(0.12) 0 (0.17) 0.03 (0.11) 0
Constant −7.825* −2.48 3.279 0.69 1.49 0.46
(3.15) 0.04 (4.73) 0.51 (3.21) 0.66
R2 (overall) 0.906 0.678 0.697
Endogeneity 0.2383 0.2533 0.2664
DV: urban employment
GDP 0.661*** 6.32 0.392** 4.76 0.480*** 5.87
(0.1) 0 (0.08) 0 (0.08) 0
Constant −2.909 −1.02 4.577 1.95 2.068 0.9
(2.87) 0.34 (2.35) 0.09 (2.29) 0.4
R2 (overall) 0.757 0.755 0.706
Endogeneity 0.3516 0.2696 0.198
DV: female employment
GDP 1.082** 3.5 0.471** 4.66 0.526*** 7.37
(0.31) 0.01 (0.1) 0 (0.07) 0
Constant −15.555 −1.84 1.217 0.42 −0.333 −0.17
(8.47) 0.11 (2.87) 0.68 (2) 0.87
R2 (overall) 0.898 0.727 0.693
Endogeneity 0.9333 0.6797 0.8102
Observations 24 32 56
Groups 8

b = beta, se = standard error, t = t-statistic, p = p-value, DV = dependent variable. *p < 0.05, **p < 0.01,

***p < 0.001.

Foley (2013), we use gross investment by sector as an instrument for GDP.17 We use
a two-stage generalized method of moments (GMM) estimation with sectoral fixed
effects that is robust to arbitrary homoscedasticity and clustering by sector. The
correspondingly robust first-stage F-statistic (Kleibergen–Paap Wald)18 that tests for
weak identification is 46.731 and is greater than the Stock Yogo critical value of
16.38 at the 5% significance level, suggesting the equation is not weakly identified
(see Stock and Yogo, 2005).19 We then run a C-statistic-type test of endogeneity
(Hayashi, 2000) for each of the estimations in Table  5 and report the p-values in

17
  Basu and Foley (2013) also use aggregate consumption expenditure and government expenditure as
instruments for GDP. However the Central Statistical Organisation of India does not provide estimates of
aggregate consumption expenditure and government expenditure at the sectoral level and these instruments
could not be used.
18
 See Kleibergen and Paap (2006).
19
  According to Staiger and Stock (1997), an instrument is weak if its first-stage test statistic is less than
10. Thus using this second decision rule we can again conclude that the instrument is not weak.
 Jobless growth in India   857
the row titled ‘Endogeneity’. The null hypothesis is that the specific regressor can
20

be treated as exogenous and a rejection implies that the original OLS estimates are

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
not consistent. The p-values in all cases are large and greater than 0.05, so the null
hypothesis of exogeneity cannot be rejected and the OLS estimates are consistent.
In conclusion, we find strong evidence of the decline of the K-V coefficient in the
aggregate and across employment categories using estimates of total employment. We
also find a declining trend in the K-V coefficients across sectors, though the trends are
not statistically significant for every sector. Thus we are able to conclude that growth
in India has become increasingly jobless over time.

4.3  Decomposition of the Kaldor–Verdoorn effect

4.3.1  Sectoral effects 
So far we have examined K-V effects within each sector, but depending on its size and
growth rate, each sector will contribute a different share to the growth of the K-V coef-
ficient in the aggregate. Here we conduct a simple decomposition exercise that esti-
mates each sector’s weighted contribution to the growth in the K-V coefficient using
a two-sector decomposition exercise based on Basu and Foley (2013) and extend it to
incorporate n sectors. The decomposition is derived below, where E is employment, e
is the growth rate of employment, Y is output and y is growth rate of output. The K-V
coefficient in the aggregate (de/dy) is derived as the product of the sector-specific K-V
coefficients as estimated in the previous section and the relative growth rate of a sector
weighted by its share in total employment:
For sectors i to n, let Si = Ei/E be the share of sector i in total employment

Then e = ∑Si * ei
Differentiate this expression with respect to y
de / dy = ∑Si (dei / dy ) if the sectoral share of employment is constant over time (4)
de / dy = ∑ S i ( dei / dyi ) * ( d yi / dy )
≅ ∆ e / ∆ y = ∑Si ( β i * ( ∆ yi / ∆ y ))

The last step is an approximation as differentials over continuious time are replaced
with differences over discrete time. Figure  4 presents the sectoral contribution to the
aggregate K-V coefficient for the subperiods 1978–94 and 1994–2010. Each sector’s
absolute contribution to the aggregate K-V effect is normalised by the total to obtain its
contribution in percentage terms. Clearly the contribution of agriculture to the aggre-
gate K-V coefficient has fallen sharply between the two periods, as has the contribution
of other services. On the other hand, construction, trade, hotels and restaurants and
manufacturing contribute relatively higher shares to the K-V coefficient over time. Again
the high contribution of construction is noteworthy, especially in the context of the rural
employment guarantee and its focus on public works, though construction work in India
continues to be poorly paid, low skilled and extremely hazardous in nature.
The question then becomes: given that sectoral K-V coefficients have been flat or
declining over the period, how has the relative growth of sectors constituted these trends
at the aggregate level? As seen in Figure 5, the relative growth of agriculture has been

  The test is run using the ivreg2 module in Stata with the endog option.
20
858  S. Tejani

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Fig. 4.  Sectoral contribution to aggregate K-V coefficient.
Source: Author’s illustration.

Fig. 5.  Relative growth by sector, 1983–2010.

Source: Author’s illustration.

consistently and far below average, with a sharp dip beginning in the early 1990s, while
the construction, trade, hotels and restaurants and transport and communications sec-
tors have accelerated their growth as compared with the average. However, the sectors
that are growing relatively rapidly comprise a much smaller share of total employment
 Jobless growth in India   859
as compared with agriculture and, combined with falling sector-specific K-V coefficients
over the period, constitute a trend of falling aggregate employment elasticity over time.

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Again, this shows that the low relative growth of agriculture and its large weight in employ-
ment have been important factors in the decline in K-V coefficients in the aggregate.

4.3.2 Services sector
There is considerable heterogeneity in the various service subsectors in India, with
some characterised by large proportions of informal labour or self-employment and
low productivity while others employ highly skilled labour and display high productiv-
ity. For illustrative purposes, we decompose the aggregate K-V coefficient for services
into nine subsectors at the one-digit level21 of the NIC (2004) for the period 2000–10
using the same decomposition exercise as in Section 4.3.1 (equation 4). Table 6 pre-
sents both terms of the decomposition exercise for the years 2005 and 2010: the K-V
coefficients and the relative growth for the various subsectors.
As evident from Table 6, the K-V coefficient for the aggregate service sector halved
between the years 2005 and 2010 (from 0.34 to 0.15) and declined for all subsectors
except storage and public administration and defence, which displayed negative coef-
ficients in 2005. Some of the largest declines came in financial intermediation, hotels
and restaurants, other services and real estate.
Further, in 2010, financial intermediation, communications and storage either accel-
erated their relative growth rates or maintained their higher-than-average growth rates
while growth in hotels and restaurants, wholesale and retail trade, transport and other
services fell to below average levels. As a result, the contribution of high-productivity
FIRE industries to the aggregate service K-V coefficient increased despite a fall in their
K-V coefficient, though output and thus labour productivity in these sectors remain
difficult to define and measure (Figure 6).22 The share of FIRE industries in total ser-
vice employment however is relatively small (8.3% in 2010). The higher contribution
of public administration and defence to the aggregate service sector K-V coefficient is
also due to its higher relative growth. Employment growth in the subsector was how-
ever negative in 2005, which led to a rise in the K-V coefficient in 2010 due to the low
base effect, although employment levels in that year were lower than in 2000. Despite
their relative slowdown and declining K-V coefficients, the wholesale and retail trade
and transport subsectors together contributed more than 45% to the aggregate service
sector K-V coefficient due to their higher employment shares. Both are relatively low-
productivity services with a large number of self-employed and informal workers. The
contribution of other services,23 also a low-productivity sector, declined due to falling
K-V coefficients as well as lower growth over the period.
In conclusion, we see that growth in services has accelerated in relatively higher pro-
ductivity subsectors and that every unit of output growth in services has generated lower
21
  Except for the transport, storage and communications sector, which is decomposed at the two-digit
level of NIC (2004) into transport, storage and communications industries, respectively. Education, health
and social work, other community, social and personal activities and activities of private households as
employers (sections M, N, O and P of the NIC, 2004) are combined as ‘other services’ as disaggregated
GDP data at factor cost are not available separately for each. Further, in this case annual GDP data are
matched with employment data as the former are not available as a quarterly series at this level of detail.
22
  Basu and Foley (2013) suggest that employment in service industries has a lower responsiveness to out-
put because of measurement problems and the weak link between aggregate demand and income in the FIRE
industries. In general, however, services contribute a higher share to the K-V coefficient in the Indian case.
23
  Includes education, health and social work, other community, social and personal activities and activi-
ties of private households as employers.
860  S. Tejani
Table 6.  Service sector decomposition

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
K-V coefficient Relative growth

Sector 2005 2010 2005 2010

Wholesale and retail trade 0.298 0.110 1.156 0.900

Hotels and restaurants 0.493 0.223 1.199 0.780
Transport 0.380 0.308 1.298 0.742
Storage −1.858 1.678 0.474 0.996
Communications 0.119 0.015 4.535 3.359
Financial intermediation 0.932 0.305 0.757 1.682
Real estate, renting and business activities 1.601 0.596 0.909 0.898
Public admin and defence −1.210 0.219 0.402 0.961
Other services 0.462 0.081 0.925 0.600
Total 0.338 0.153

Fig. 6.  Sectoral contribution to services K-V coefficient.

Source: Author’s illustration.

growth in service employment over time. These high productivity subsectors comprise a
smaller share of total employment and thus contribute relatively less to the aggregate ser-
vice sector K-V coefficient than the subsectors that have decelerated during the period.
Thus most employment growth in the services still remains concentrated in relatively
low-productivity subsectors, even as their K-V coefficients decline over time.

5.  Some explanations for jobless growth

We have found that K-V effects have cut into employment growth in the Indian context at
the aggregate and sectoral levels, but the question remains: what explains these K-V effects
 Jobless growth in India   861
and their particular acceleration in the period between 1994 and 2010? The consensus in
the literature is that rising labour productivity is a reflection of increasing capital intensity

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
in the economy, though there is disagreement on why capital intensity has risen particularly
fast in the post-reform period. Figure 7 displays capital intensity, labour productivity and the
output–capital ratio for the aggregate economy in India based on 2005 purchasing power
parities using data from the Extended Penn World Tables (Marquetti and Foley, 2011).
It is clear from the figure that capital intensity displays an almost exponential growth
starting in the mid- to late-1990s, with a corresponding rise in labour productivity.
Although the level of capital intensity in India is much lower than in countries such as
China, Malaysia or even Thailand, its output–capital ratio is comparable to countries
that have a much higher level of income per capita.
Some of the competing explanations in the literature for rapid labour productivity
growth (or higher capital intensity) in India are as follows: (i) relative factor costs
are skewed in favour of capital and have led producers to choose more capital-
intensive techniques in order to control rising wage costs (Goldar, 2009); (ii) the
composition of domestic and export demand has shifted to products and services
that require a greater use of technology, which is itself the result of the distribution
of income away from wages to profits and a ‘demonstration effect’ (Patnaik, 2009;
Chandrashekhar, 2009); and (iii) rising international competition has led producers
to adopt frontier technologies to cater to foreign demand but also to compete with
rising import penetration in the domestic economy (Dasgupta and Singh, 2005;
Chandrasekhar and Ghosh, 2007). We examine each of these explanations in turn.

5.1  Relative factor costs

If relative factor costs have encouraged the use of capital-intensive technologies, as
argued by Goldar (2009), strongly rising wages would be driving the rise of labour

Fig. 7.  Labour productivity and capital–output ratio.

Source: Author’s illustration based on Marquetti and Foley (2011).
862  S. Tejani
productivity in the aggregate. In this subsection we present some descriptive statistics
on wages and conduct a preliminary statistical exercise by regressing labour productiv-

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
ity on wages. If wage costs are not positively and significantly correlated with the rise in
productivity, then clearly there must be another reason for the rise in capital intensity.
Figure 8 presents daily average wage/salary earnings of regular employees between
the ages of 15 and 59 years from 1994–95 to 2009–10, adjusted for inflation using the
Consumer Price Index (CPI) for Industrial Workers (CPI-IW). Although the wages of
urban males have inched up by an average of about 4% per year during the period, the
wages of rural male workers have been more or less stagnant since the year 2000. More
surprisingly, female wages in the rural areas have shown a sharp decline since the year
2000. With such stagnation in wages at the aggregate level, it is difficult to imagine that
high wage costs are incentivising producers to shift to more capital-intensive produc-
tion. Although we look at regular salaries and wages here, the conditions for casual and
informal labour with respect to wages and working conditions are much worse.
As a preliminary investigation, we construct a sectoral panel of labour productivity
and daily average salary earnings of regular employees by the following categories:
urban male, urban female, rural male and rural female. Ostensibly, it is these high sala-
ries of regular employees, and not the daily wages of casual labour, that are prompting
the shift to more capital-intensive production.
Salaries and wages are deflated to 2004–05 dollars using the CPI for agricultural
workers (CPI-AL) and CPI-IW for the relevant category of workers to maintain con-
sistency with labour productivity estimates.24 Because labour markets in India are
highly segmented and the level of wages differs greatly for each category of workers,
we retain them as separate variables in our panel instead of deriving a single national
average wage for each sector. Controlling for sector through the addition of sectoral

Fig. 8.  Average daily salaries and wages for regular employees.

Source: Author’s illustration based on NSSO (various years).

24
  CPI-AL and CPI-IW are obtained from MoF (2011).
 Jobless growth in India   863
fixed effects, we regress productivity on all four wage categories using robust standard
errors.

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Table 7 presents the results of the exercise: none of the wage categories is correlated
with labour productivity at standard levels of significance. On the other hand, when we
regress each of the wage categories on labour productivity separately, the coefficient
on productivity is positive and statistically significant. This suggests that rising produc-
tivity in fact has an impact on wages, though the size of the coefficient will vary when
other controls are added to a standard wage equation. One possible reason for such
an effect is the fact that wage costs as a proportion of total costs will be lower as value
added increases, allowing more room for wage increases.

5.2  Shift in the composition of demand

We now investigate if a shift in the composition of demand has contributed to a rise in
labour productivity in India. If, as Patnaik (2009) and Chandrashekhar (2009) argue,
demand has shifted to products that require greater use of technology, a larger pro-
portion of output would be accounted for by high-productivity sectors and aggre-
gate productivity would be higher than it would be in the absence of such a shift. We
decompose aggregate labour productivity into an average of the difference between
weighted sectoral output and employment growth and a ‘reallocation effect’ (Ocampo
et al., 2009). The latter can be interpreted as a shift in demand between sectors; thus
if output is reallocated to a sector that has a higher share of output than employment,
it will raise aggregate productivity (Ocampo et al., 2009, p. 54). Conversely, if its share
of output is lower than its employment share, output growth in that sector will reduce
productivity overall. A brief description of these exercises is reproduced below, where
E is employment, Y is output and Q (= Y/E) is productivity:
For sectors {i…n}, at times 0 and 1

E0 = ∑E
i
i
0
and Y0 = ∑Y
i
0
i

Table 7. Average wages/salaries and labour productivity

Variable b/se t/p

Rural male 0.19 0.86

(0.22) 0.42
Rural female 0.019 0.33
(0.06) 0.75
Urban male 0.783 2.11
(0.37) 0.07
Urban female 0.067 0.27
(0.25) 0.8
Constant 6.251** 5.03
(1.24) 0
R2 0.681
Observations 32
Groups 8

b = beta, se = standard error, t = t-statistic, p = p-value. *p < 0.05, **p < 0.01, ***p < 0.001.
864  S. Tejani
S0 = E0i / E0

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
θ0 = Y0i / Y0


E = ( E1 − E0 ) / E0


Y = (Y1 − Y0 ) / Y0

ξ = (Q1 − Q0 ) / Q0

Log differentiating Q with respect to time, the growth rate of labour productivity is:

−1

∑(θ
  
ξ = (1 + E ) i
0
Y i − S0i E i )
i

−1

where (1 + E ) is the ‘interaction’ effect.

Adding and subtracting S0i .Y i on the right-hand side:

−1
  i  i
∑[S
  i
ξ = (1 + E ) i
0  Y − E  + (θ0 − S0 )Y ] (5)
i i


An equivalent expression is obtained by adding and subtracting θ 0i . E i :

−1
  i  i
∑[θ
  i
ξ = (1 + E ) i
0  Y − E  + (θ0 − S0 ) E ] (6)
i i

We can construct a counterfactual to test our hypothesis of whether a shift in the

composition of demand has contributed to a rise in labour productivity by estimating
aggregate productivity in the absence of the reallocation effects in equation (5). If this
counterfactual productivity is lower than actual productivity, it implies that the shift
of demand across sectors has contributed to this difference. Figure 9 displays the esti-
mated productivity growth and counterfactual productivity growth and we find that
the latter is significantly lower for each year in our sample.
Figure 10A presents sectoral contributions to aggregate productivity growth (or the
excess of output growth over employment growth) by year from 1983 to 2010 based
on the above decomposition. It is clear from the figure that the other services sector25
has contributed the principal share to productivity growth since 1983 followed by
the manufacturing and trade, hotels and restaurants sectors. Although Kaldor empha-
sised the manufacturing sector as being central in terms of increasing returns and
dynamic economies of scale, in India the services sector has consistently contributed
more to aggregate productivity growth than manufacturing.26 However the sectoral

  Note that ‘other services’ here includes FIRE industries and community, social and personal services.
25

  Roncolato and Kucera (2013) report the same finding.

26
 Jobless growth in India   865

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Fig. 9.  Productivity growth and its counterfactual.
Source: Author’s illustration.

contribution to productivity has become more evenly distributed since 2005 and this
might be a preliminary indication that K-V effects in the economy are becoming more
widespread over time. Figure 10B presents the sectoral contribution to productivity
growth if no reallocation effects were present, i.e. if there had been no shifts in output
across sectors for each year in the sample. The composition of productivity is quite
different with agriculture, which has a much lower level of productivity, contributing a
much larger share to aggregate productivity growth.
It is clear that the shift of demand from relatively low productivity agriculture to the
higher-productivity services sector has contributed to higher aggregate productivity
growth. In other words, in terms of the decomposition exercise, output has been con-
sistently reallocated to sectors in which the output share already exceeds the employ-
ment share, thus pushing up productivity growth in the aggregate.

5.3  International competition

Dasgupta and Singh (2005) speculate that greater intensity of competition at the
domestic and international levels in developing countries such as India might be driv-
ing the exceptional growth of labour productivity, which then compromises employ-
ment growth. At the domestic level, it is possible that rising import competition due
to liberalisation leads firms to adopt frontier technologies and produce better quality
goods that compare well with similar foreign goods in the market. At the international
level, greater export orientation of production in the Indian economy could also be pro-
moting productivity gains through increasing export competition. We attempt to test
this hypothesis in the following way: we construct an import penetration index, export
propensity index and overall trade dependence index by sector using input–output
data for the Indian economy from the early 1990s to the mid-2000s from the OECD
(2011). If competition is indeed one of the factors contributing to rapid productivity
growth, the sectors that show large gains in productivity will also display rapidly rising
866  S. Tejani
(a) Actual

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
(b) Counterfactual

Fig. 10.  Sectoral contribution to productivity growth. (A) Actual and (B) counterfactual.
Source: Author’s illustration.
 Jobless growth in India   867
levels of trade dependence. The indices are based on Mikic and Gilbert (2009) and
constructed as follows, where X is exports, M is imports, VA is value added, i is sector,

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
d is country and s is the set of all other countries:27

Import Penetration Index: I i =

∑ sMisd
VAid − ∑ sX ids + ∑ sMisd

Export Propensity Index: E i =

∑X s ds
*100
VAid

Trade Dependence Index: Ti =

∑X s ids
+ ∑ sMisd
VAid
The exercise produces mixed results that are displayed in Table  8 (sorted by trade
dependence). The mining and quarrying sector, which shows the highest growth in
trade dependence, also displays high growth in productivity in terms of its compound
annual growth. Other services, which includes FIRE industries and community, social
and personal services, displays the highest growth in productivity over the period. The
FIRE industries show a large increase in export propensity over the period, but this
effect is dampened when it is subsumed under other services. Agriculture and allied
activities show a low growth of both trade penetration and productivity. The transport,
storage and communications and trade, hotels and restaurants sectors appear to have
raised productivity despite a contraction in their levels of trade penetration and appear
to contradict the hypothesis. The utilities and construction sectors have zero or near-
zero levels of trade dependence and are not reported here.
Overall, we find that some sectors displaying rapid productivity growth during the
period 1994–2005, such as mining, manufacturing and other services, have also expe-
rienced rising competition through a fairly high degree of import penetration, export
growth or both. However, rising competition is clearly not the only factor responsible
for productivity growth as is evident in the case of the utilities sector, which experi-
enced the highest growth of productivity—7.1% compound annual growth rate—over
the period and had almost zero trade dependence. The endogenous element of pro-
ductivity growth as a result of the growth of output remains important.

6.  Concluding remarks

Using a Kaldorian framework of endogenous productivity growth, we find that K-V
effects in India have become more predominant over time, especially in the post-reform
(1994–2010) period. Our estimated K-V coefficients for both formal sector and total
employment have dropped dramatically, suggesting that India has leapfrogged into a
high-productivity regime without the broad-based expansion of labour-intensive pro-
duction that has been characteristic of fast-growing economies in East Asia. India, on
the other hand, stands out for its rapid expansion of high-productivity service sectors.
Labour productivity growth in India has been a double-edged phenomenon that has
hurt employment growth even as it has remained a critical factor in the expansion
of output. A caveat to our results are known weaknesses of the estimation of services
GDP in India, especially with respect to the output of the private corporate sector

27
 The export propensity and trade dependence indices have been modified from Mikic and Gilbert
(2009) and estimated at the sectoral level.
868  S. Tejani
Table 8.  Trade and productivity growth, 1994–2005

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Sector Growth of index Productivity
CAGR (compound
Import Export Trade annual growth
penetration propensity dependence rate)

Mining and quarrying 26.9 33.06 150.3 5.26

Manufacturing 15.97 14.96 35.92 3.50
Other services −1.53 7.16 4.1 5.77
(a) FIRE 1.65 24.45 24.71 –
(b) Community, social and −4.82 −1.19 −6.12 –
personal services
Agriculture, hunting, forestry 0.79 1.02 1.8 1.67
and fishing
Transport, storage and −0.25 −1.2 −1.39 3.96
communications
Trade, hotels and restaurants 0.17 −3.22 −3.06 3.24

and the informal segment of the tertiary sector (see Tendulkar, 2007; Shetty, 2007).
Although these weaknesses will be reflected in the estimations presented here, they will
affect the results no more than other studies that utilise the same data sources.
We examine some explanations for why these K-V effects have become pronounced
over time and are not convinced that wage pressure has been one of the reasons for the
rising labour productivity and capital intensity of output. A shift in the composition
of demand towards higher-productivity sectors does however appear to be an impor-
tant part of the explanation. We also find mixed evidence that forces of international
competition have generated pressures to adopt more capital-intensive techniques of
production.

Bibliography
Basu, D. and Foley, D. K. 2013. Dynamics of output and employment in the US economy,
Cambridge Journal of Economics, vol. 37, no. 5, 1077–1106
Besley, T. and Burgess, R. 2004. Can labor regulation hinder economic performance? Evidence
from India, Quarterly Journal of Economics, vol. 119, no. 1, 91–134
Bhalotra, S. R. 1998. The puzzle of jobless growth in Indian manufacturing, Oxford Bulletin of
Economics and Statistics, vol. 60, no. 1, 5–32
Bhattacharjea, A. 2006. ‘Labour Market Regulation and Industrial Performance in India:
A Critical Review of the Empirical Evidence’, Working Paper no. 141, Centre for Development
Economics
Bhattacharjea, A. 2009. The effects of employment protection legislation on Indian manufactur-
ing, Economic and Political Weekly, vol. 44, no. 22, 55–62
Bhattacharya, B. B. and Sakthivel, S. 2004. ‘Economic Reforms and Jobless Growth in India in the
1990s’, Working Paper no. E/245/2004, Institute of Economic Growth
Bhattacharya, P. 2013. Everything you wanted to know about the Sen–Bhagwati debate, Mint,
20 July, http://www.livemint.com/Politics/zvxkjvP9KNfarGagLd5wmK/Everything-you-
wanted-to-know-about-SenBhagwati-debate.html last accessed 13 May 2015
Chandrasekhar, C.P. 2009. Product Composition, Jobless Growth and the Role of Demand-
Side Influences. Paper written for the ILO Sub-regional office, New Delhi
Chandrasekhar, C. P. and Ghosh, J. 2007. Recent employment trends in India and China: an
unfortunate convergence? Social Scientist, vol. 35, nos 3–4, 19–46
CSO. 2011a. Statement 4 & 5: Gross Domestic Product by economic activity at current and
2004-05 prices, National Accounts Statistics Back Series 2011 (1950–51 to 2004–05), Central
 Jobless growth in India   869
Statistical Organisation, India, http://mospi.nic.in/Mospi_New/upload/back_series_2011.htm
downloaded 26 June 2012

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
CSO. 2011b. Statement 10: Gross Domestic Product by Economic Activity at 2004-05 prices,
National Accounts Statistics 2011: An Overview, Central Statistical Organisation, India, http://
mospi.nic.in/Mospi_New/upload/NAS_web_2011-29aug.htm downloaded on 26 June 2012
Fallon, P. R. and Lucas, R. E. B. 1991. The impact of changes in job security regulations in India
and Zimbabwe, World Bank Economic Review, vol. 5, no. 3, 395–413
Goldar, B. 2000. Employment Growth in Organised Manufacturing in India, Economic and
Political Weekly, vol. 35, no. 14, 1191–1195
Goldar, B. 2009. ‘Impact of Trade on Employment Generation in Manufacturing in India’, Working
Paper no. E/297/2009, Institute of Economic Growth
Hayashi, F. 2000. Econometrics, Princeton, Princeton University Press
IMF. 2010. World Economic Outlook: Rebalancing Growth. Washington, DC: International
Monetary Fund
Kaldor, N. 1966. Causes of the Slow Rate of Economic Growth of the United Kingdom: An Inaugural
Lecture, Cambridge, UK, Cambridge University Press
Kaldor, N. 1967. Strategic Factors in Economic Development. Ithaca, NY, New York State School
of Industrial and Labor Relations, Cornell University
Kaldor, N. 1968. Productivity and growth in manufacturing industry: A reply, Economica, 385–
391 (November)
Kannan, K. P. and Raveendran, G. 2012. Counting and profiling the missing labour force,
Economic and Political Weekly, vol. 47, no. 6, 77–80
Kleibergen, F. and Paap, R. 2006. Generalized reduced rank tests using the singular-value
decomposition, Journal of Econometrics, vol. 127, 97–126
Knotek, E. S., II 2007. How useful is Okun’s law? Economic Review, Fourth Quarter, 73–103
McMillan, M. and Rodrik, D. 2011. Globalization, structural change and productivity growth,
in Bacchetta, M. and Jansen, M. (eds), Making Globalization Socially Sustainable, Geneva, ILO
and WTO, 49–84
Marquetti, A. and Foley, D. K. 2011. Extended Penn World Tables: Economic Growth Data Version
4.0, https://sites.google.com/a/newschool.edu/duncan-foley-homepage/home/EPWT down-
loaded on 10 August
Mikic, M. and Gilbert, J. 2009. Trade Statistics in Policymaking: A  Handbook of Commonly
Used Trade Indices and Indicators, rev. edn, Thailand, United Nations Economic and Social
Commission for Asia and the Pacific
MoF. 2011. Economic Survey 2010–2011, New Delhi, Ministry of Finance
Moreno-Monroy, A. I., Pieters, J. and Erumban, A. A. 2012. ‘Subcontracting and the Size and
Composition of the Informal Sector: Evidence from Indian Manufacturing’, IZA Discussion
Paper no. 6785
Nagaraj, R. 2000. Organised manufacturing employment, Economic and Political Weekly, vol. 35,
no. 38, 3445–3448
Nagaraj, R. 2004. Fall in organised manufacturing employment: a brief note, Economic and
Political Weekly, vol. 39, no. 30, 3387–90
Nataraj, S. 2011. The impact of trade liberalization on productivity: evidence from India’s for-
mal and informal manufacturing sectors, Journal of International Economics, vol. 85, no.  2,
292–301
NCEUS. 2009. The Challenge of Employment in India: An Informal Economy Perspective, New
Delhi, National Commission for Enterprises in the Unorganised Sector
NSSO. Various years. Employment and Unemployment Situation in India, New Delhi, National
Sample Survey Office
Ocampo, J. A., Rada, C. and Taylor, L. 2009. Growth and Policy in Developing Countries:
A Structuralist Approach, New York, Columbia University Press, 131–136
OECD. 2011. STAN Input-Output: Input Output Database, STAN: Structural Analysis
Statistics (database), http://dx.doi.org.lib-ezproxy.unog.ch/10.1787/data-00650-en down-
loaded on 2 November 2011
Okun, A. M. 1962. ‘Potential GNP: Its Measurement and Significance’, Cowles Foundation
Paper no. 190
870  S. Tejani
Papola, T. S. 2007. Employment trends in India, in Basu, K. (ed.), The Oxford Companion to
Economics in India, New Delhi, Oxford University Press

Downloaded from https://academic.oup.com/cje/article-abstract/40/3/843/2604950 by Indian Institute of Management Kozikode user on 27 September 2019
Patnaik, P. 2009. ‘A Perspective on the Growth Process in India and China’, IDEAS Working Paper
no. 05/2009
Pieper, U. 2003. Sectoral regularities of productivity growth in developing countries: a Kaldorian
interpretation, Cambridge Journal of Economics, vol. 27, no. 6, 831–50
Rani, U. and Unni, J. 2004. Unorganised and organised manufacturing in India: potential for
employment generating growth, Economic and Political Weekly, vol. 39, no. 41, 4568–80
Roncolato, L. and Kucera, D. 2013. Structural drivers of productivity and employment growth:
a decomposition analysis for 81 countries, Cambridge Journal of Economics, vol. 38, no.  2,
399–424
Rowthorn, R. E. 1975. A reply to Lord Kaldor’s comment, Economic Journal, vol. 85, no. 340,
897–901
Shetty, S. L. 2007. Status paper on database issues of the service sector, Economic and Political
Weekly, vol. 42, no. 37, 3723–3726
Dasgupta, S. and Singh, A. 2005. ‘Will Services be the New Engine of Economic Growth in India?’
Working Paper no. 310, University of Cambridge Centre for Business Research
Skott, P. 1999. Kaldor’s growth laws and the principle of cumulative causation, in Setterfield,
M. (ed.), Growth, Inflation and Unemployment: Essays in Honor of John Cornwall, London,
Macmillan, 166–179
Staiger, D. and Stock, J. H. 1997. Instrumental variables regression with weak instruments,
Econometrica, vol. 65, 557–86
Stock, J. H. and Yogo, M. 2005. Testing for weak instruments in linear IV regression, in Andrews,
D. W. K. and Stock, J. H. (eds), Identification and Inference for Econometric Models: Essays in
Honor of Thomas Rothenberg, Cambridge, UK, Cambridge University Press, 80–108
Sundaram, K. and Tendulkar, S. D. 2002. ‘The Working Poor in India: Employment–Poverty
Linkages and Employment Policy Options’, Issues in Employment and Poverty Discussion
Paper no. 4, International Labour Organization
Tendulkar, S. D. 2007. India’s growing services sector: Database problems and issues, Economic
and Political Weekly, vol. 42, no. 37, 3721–3722
Unni, J. and Bali, N. 2002. Sub-contracted Women Workers in the Garment Industry in India, in
Radhika Balakrishnan (ed.),The Hidden Assembly Line: Gender Dynamics of Sub-contracted Work
in a Global Economy, Kumarian Press, Connecticut
UNSD. 2011. Demographic Statistics, United Nations Statistics Division, http://data.un.org/
Data.aspx?d=POP&f=tableCode%3a22 [date last accessed: 4 April 2011]
Verdoorn, P. J. 1949. Factors that determine the growth of labour productivity, in McCombie, J.
S. L., Pugno, M. and Soro, B. (eds), Productivity Growth and Economic Performance: Essays on
Verdoorn’s Law, London, Palgrave Macmillan, 28–36
World Bank. 2007. World Development Report 2008: Agriculture for Development, Washington, DC,
World Bank
World Bank. 2011. Perspectives on Poverty in India: Stylized Facts from Survey Data, Washington,
DC, World Bank
Young, A. 1928. Increasing returns and economic progress, Economic Journal, vol. 38, 527–42