Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
WP. No.: SAUFE-WP-2020-005
Does City Size Matter for Migration and Poverty: A study of Million-plus cities in India
Arup Mitra Professor
Faculty of Economics, South Asian University Akbar Bhawan, Chanakyapuri New Delhi 110021, INDIA Email: [email protected]
Sabyasachi Tripathi Postdoctoral Research Fellow
National Research University Higher School of Economics Moscow, RUSSIA
Email: [email protected]
Working Paper Number: SAUFE-WP-2020-005
http://www.sau.int/fe-wp/wp005.pdf
FACULTY OF ECONOMICS SOUTH ASIAN UNIVERSITY
NEW DELHI
March, 2020
1
Does City Size Matter for Migration and Poverty:
A study of Million-plus cities in India
Arup Mitra1
Sabyasachi Tripathi2
Abstract
City size plays an important role in the context of growing urbanization in developing countries
such as India. Are economic activities concentrated in a few large cities? What are the important
economic factors that determine the size of large cities? Is it possible to trace different groups
among the large cities? Is the relationship pertaining to city size, migration, and poverty
quantifiable? In this paper, we answer these questions by considering 47 million-plus cities in
India. The estimated Pareto distribution suggests that within the group of large million-plus cities
production is not so diversified though the extreme form of concentration like primacy or rank
size rule does not hold. In fact, three broad groups are evident as brought out by the results of the
cluster analysis. Both OLS and Tobit regression results indicate that the wage rate for which the
is the average monthly per capita expenditure (MPCE), amount of municipal solid waste
generation (proxy for the quantum of economic activities) and the number of salaried employed
persons generate a positive effect on city population while the number of registered motor
vehicles, a proxy for congestion, has a negative effect on it. However, the infrastructure index
calculated through factor analysis does not show any effect on the city size. Based on the results
of factor analysis increased migration and reduced poverty are seen to be associated with city
size. Finally, we suggest that to optimize the benefits of future urbanization, the promotion of
non-million plus cities and towns is essential with higher infrastructural investment. Small cities
and towns will be the next destination for increasing the rural to urban migration and reduction
of rural poverty in India.
Keywords: city size, agglomeration economies, migration, poverty, congestion
1Professor of Economics, South Asian University, New Delhi, India, email: [email protected] 2Postdoctoral Research Fellow, National Research University Higher School of Economics,
Moscow, Russia, email: [email protected]
2
1. Introduction
City size captures a central position in the urban economics literature. Every city can be said to
be in equilibrium while at the same time city size differs significantly within a given country.
Even within a specific category, say million plus cities, population size may differ substantially
in a large country like India where it is difficult to have the rule of primate city. Even if primacy
were to hold, different regions would warrant its existence, thus making impossible for one city
to dominate the whole nation. Thus, it may be pertinent to revisit the set of million plus cities and
examine the variations in their sizes explainable in terms of the nature of activities or the
structure of the cities, wages and so on. The wages are likely to differ significantly as the scale
economies and productivity vary across space. Though the large cities with their higher scale
economies also encounter higher diseconomies and costs, the benefits do not get completely
swept away as the cost reduction mechanisms are plenty, implying that real wages may show a
positive association with city size. Though the data availability compels us in some sense to
restrict the analysis to the million plus cities, there is sufficient reason to consider them on
analytical grounds as well. Large cities are a good proxy for a city district as they cover a large
portion of a district than smaller cities and further, Indian urbanization is mainly concentrated in
and around the large cities: 47 million-plus cities in India accommodated about 40 percent of the
total urban population in 2011.
The paper begins by examining the size variability within the million plus category by adhering
to the Pareto Distribution framework. If the coefficient of city size is seen to be greater than 1
then primacy may hold even within the category of the million plus cities. On the other hand, if
the coefficient is unity it implies rank size rule and the lower magnitude of the coefficient is
indicative of dispersion rather than concentration of economic activities. Secondly it delves into
the issue of differences in the structure and wages across cities which are considered to be the
major determinants of city size. A cluster analysis confirms that within the category of million
plus cities different groups are traceable. Extending the analysis further it is argued that rural to
urban migration is a significant constituent of population increase in cities. If so, what are the
drivers of population movement and whether city poverty is related to the population shift from
the rural areas? Migration to large cities may help poor escape their vulnerability; hence
provision for space selective migration is beneficial for a country’s poverty reduction objective.
3
Even when economic growth may not be spectacular cities are expected to be the better
performers, which would result in higher mean effect in poverty reduction in the poverty
decomposition framework [Mazumdar and Son, 2002 and Bhanumurthy and Mitra, 2004].
The paper is organized as follows. In the next section, we review the literature in brief.
Distribution and determinants of city sizes are explained in sections 3 and 4, respectively. The
relationships among city, migration, and poverty are explained in section 5, followed by major
conclusions and policy implications in section 6.
2. Review of literature
One of the major sources of agglomeration economies is related to the infrastructure endowment.
A classic paper by Tiebout (1956) argued that the availability and quality of public facilities and
services, such as schools, municipal golf courses, beaches, parks, police protection, roads, and
parking facilities, affect the decision-making process for choosing a municipality. Harris and
Todaro (1970) explained that rural–urban migration mainly occurs as a result of expected income
differentials between rural and urban areas: this implies that the urban areas with much better
conditions would attract migrants on a large scale.
The India Infrastructure Report 2011 (Infrastructure Development Finance Company 2011)
noted that inadequate access, poor quality, and poor reliability are the major problems with urban
water supply. Nair (2012) estimated that approximately 21% of the urban population lives in
squatter settlements with poor access to basic services: about 30–50% of the households do not
have sewerage connections, and less than 30% of the total wastewater is treated. Toutain and
Gopiprasad (2006) found inadequate urban service provisioning mainly in terms of drinking
water, sanitation, energy, transport, solid waste management, environmental degradation, and
pollution. Very importantly, findings from the report on the Urban Infrastructure in India
(Federation of Indian Chambers of Commerce & Industry 2011) confirm that present urban
infrastructure is grossly inadequate to meet the demand of the existing urban population. Pradhan
(2007) investigated the impact of infrastructure on urbanization, using a composite infrastructure
development index and confirmed that it has a significant positive impact. However, Tripathi
(2018) suggested that improvement in infrastructure does not increase population agglomeration
4
(measured by size, density, and growth rate of city population) in large cities, though it
substantially improves the potential contribution of the cities to national economic growth.
Whether growth benefits spill-over in terms of wellbeing improvements is an important question.
As employment is the basic link between the two, we may turn to this aspect first. Several
studies (e.g., Mehrotra et al. 2014; Maiti, 2015; Institute for Human Development, 2014; Bhalla
and Kaur, 2011; Papola and Sahu, 2012) have tried to understand the trends and patterns of
employment and unemployment in India though few of them reflect specifically on urban
employment. Chen and Raveendran (2012) explore the trends in urban employment, with a
special focus on urban informal employment. Ramaswamy and Agrawal (2012) found that
manufacturing employment in urban India grew at a rate (2.8 per cent) faster than the all-India
average (1.8 per cent) over the period 1999–2000 through 2009–10. Indian Institute for Human
Settlements (2012) using data from Economic Census, noted that workforce participation rates
are highest in the “major metros” (population with 4 million plus), and employment in “high-
tech” sector (ICT, high technology manufacturing, and fast growing exports) is also highly
concentrated in the large cities. Mitra and Singh (2019) estimated that with increased
urbanization the unemployment rate rises as the possibility of disguised unemployment in the
urban labour force is low compared to the rural areas. Besides, frictional unemployment and
technological unemployment are also part of the urban labour force in the relatively skilled jobs.
The India: Urban Poverty Report 2009 by the Government of India (2009) noted that across the
Indian states, poverty is negatively correlated with the level of urbanization, and large and
medium cities have a lower incidence of poverty than small cities in India. A World Bank study
(World Bank, 2010) observed that poverty is more widespread in very small towns than in large
cities. An important study by Gangopadhyay, et al. (2010) applied the small area estimation
methodology in three states of India in 2004-05, and confirmed that in the states of West Bengal,
Orissa and Andhra Pradesh, the poverty level in large cities is much lower than that in small
towns. Bhanumurthy and Mitra (2004) decomposed changes in poverty into growth effect,
inequality effect, and migration effect for two periods: from 1983 to1993/94 and 1993/94 to
1999/2000. As per their findings, rural-to urban migration contributed to poverty reduction in at
least seven of the fifteen major states and at the all-India level too, in both the periods. In other
words, the overall incidence of poverty in these states fell, though rural-urban migration might
5
have raised the incidence of poverty in the urban areas. However, Tripathi (2013b) argued that
higher levels of city economic growth and large city population agglomerations are associated
with a reduction in the poverty in cities.
3. Distribution of City Size
The framework most commonly employed to study city/town sizes is based on Pareto
distribution: G(x) = A x-a
where, G(x) is the number of urban areas with at least x people and A and a are the parameters to
be estimated from the data1.
Pareto distribution can be estimated by computing the least-squares regression:
ln G(x) = ln A – a ln x
The million plus cities were grouped into 16 different size classes and the log of the number of
cities in each size class has been regressed on the log of the average population size. Table 1
shows that while the adjusted R2 is high, the coefficient of the log of population size is negative
as expected but the magnitude is far below unity2, indicating that the rank size rule does not
apply to the million plus cities in India. In other words, population seems to be more equitably
Table 1: Impact of city population size on number of cities in each size class
Variables Dependent variable :
Log of number of cities
Log of population size of cities -0.642***
(0.114)
Constant 10.51***
(1.73)
F statistics 31.76***
R2 0.694
Adjusted R2 0.672
No. of observations 16
Standard errors in parentheses
*** p<0.01
1 If a is estimated to be 1, G(x)=Ax-1
which is known as rank size rule:
putting G(x) = 1, x=A (A is the population of the largest urban area). If A is estimated to be 1, xG(x) = A that is the
product of an urban area’s rank and population is a constant which is equal to the population of the largest urban
area. So the rank size rule implies that the second largest city is half of the largest city size and so on because rank of
the city multiplied by the population size is said to be equal to the population of the largest city. 2 We notice from the results that the sign of ‘a’ is negative and the magnitude is 0.64 (below unity) and the
corresponding t is statistically significant.
6
distributed across these cities than being concentrated in a few of the million plus cities. From
this it may be inferred that economic activities are undertaken in a number of large cities than
being pursued in a concentrated manner in a few large cities. Thus, each has its own importance
though from another perspective it may be pointed out that some of the large cities could have
been even larger (Mathur, 2019). In this sense there is a population deficit in some of the very
large cities which could have been augmented had there been greater concentration of activities.
In other words, the full benefits of the agglomeration economies have not been realized in some
of the very large cities as they still do not seem to have the double the population size of the next
tier cities.
4. Determining City Size
In the context of the determinants of city size, Mera (1973), Calem and Carlino (1991), Moomaw
(1983), and Seitz (1993) considered both agglomeration and public infrastructure simultaneously
in the analysis of productivity. For instance, Mera (1973) highlighted the importance of public
infrastructure in achieving agglomeration economies and included the level of employment as a
measure of localization economies. O'Clery and Lora (2016) observed that formal employment
depends on city size. They showed that a level of agglomeration equivalent to between 45 and 75
minutes of commuting time, corresponding to cities between 62 and 43, maximizes the impact
that the availability of skills has on the ability of agglomerations to generate formal employment
in Colombia. The studies on the relationship between city size and urban wages claimed that
productivity and wages are higher in larger cities. Echeverri-Carroll and Ayala (2009) estimated
that a doubling of the human capital density in a metropolitan area results in approximately a 2
percent increase in average individual hourly wages in the United States. A simple model by
Haynes (1973) however, explained that large cities have higher crime rates per person than small
cities. Krugman (1991) pointed out that spatial economic structure is dispersed due to centrifugal
forces which include pure external diseconomies.
In the context of India, Pandey (1977) examined the role of socio-economic variables in
determining the rate of urbanization at the state level. The author noted a positive effect of
industrialization, negative effect of cropping intensity, and no effect of average worker income
on the urbanization rate. Mathur (2005) urged that post-liberalization urban growth was driven
by the substantial growth of the urban population and changes in the share of employment in the
7
manufacturing and service sectors. Sridhar (2010) estimated the determinants of city growth and
output both at the district and city levels and showed that factors such as proximity to a large city
and the process of moving from agriculture to manufacturing, determines the size of a city.
Tripathi (2013a) investigated the economic determinants of the population size of the large cities
(750,000 or more inhabitants) and the external diseconomies such as city vehicle density were
seen to have a negative effect on the size of the city population. However, geographical,
environmental, second nature geography, infrastructure, and government policies also matter for
the population size.
Based on the literature review, infrastructure, city structure, employment type, wage rate and
diseconomies such as crime rate appear to be some of the major determinants of city size. City
structure can be envisaged in terms of solid waste generation and registered motor vehicles. An
industrial city vis-à-vis a services led city would generate different types of solid waste.
Similarly certain activities may involve motor vehicles to a larger extent in comparison to the
other activities. The nature of occupation can be taken in terms of employment status and the
wage rate, which in our analysis is captured through consumption expenditure. Since wage rate
may have an endogeneity problem consumption expenditure is possibly a good instrument.
Crime rate is a broad proxy for the diseconomies, and infrastructure involves a number of
components as mentioned below. Since there are three types of employment (self-employed,
regular salaried and casual wage) two dummies have been introduced, taking casual wage as the
reference category.
We have tried to construct an index of infrastructure, using factor analysis, for each of the
million plus cities. Generally, infrastructure is defined as a relatively permanent and foundational
capital investment of a country that facilitates smooth economic activity. It includes
administrative, telecommunications, transportation, utilities, education, health care, research and
development, and training facilities (Tripathi, 2018). To create infrastructure index, given the
availability of data, we consider city-wise different infrastructure variables, such as total road
length, total number of latrines, total water supply, total number of electricity connections, total
number of hospitals, and total number of schools, colleges, and universities, which are then
combined through factor analysis to construct the infrastructure index.
8
Before performing factor analysis, we evaluate and ensure the validity of the data. Validity refers
to the closeness of the measured values. We measure the validity using the Kaiser-Meyer-Olkin
(KMO) index and the Bartlett’s test of Sphericity. We use STATA version 13 (Stata Corp,
College Station, TX, USA) to perform the KMO test. The KMO value is 0.595, indicating that
factor analysis is relatively suitable. The significance probability of Bartlett’s test of Sphericity is
0.000 < 0.01, which rejects the original hypothesis and indicating that the parameters are related,
and thus are suitable for factor analysis. The test results are shown in Table 2.
Table 2: KMO and Bartlett’s test
KMO measure of sampling adequacy 0.595
Bartlett’s test of sphericity Approximate chi-square 72.447
Df 15
Sig. 0.000 KMO, Kaiser-Meyer-Olkin.
Table 3 presents the initial eigenvalues. An eigenvalue is the variance of the factor. The first
factor accounts for the most variances, the second account the next highest amount of variance,
and so on. Some of the eigenvalues are negative which indicates that the matrix is not full rank.
Therefore, we can consider at most three factors for the analysis. However, we retained only one
factor as the KMO criterion suggests that factors with Eigenvalues ≥ 1 are to be considered
(Appendix Fig. 1). Table 3 shows that factor1 accounts for about 83% of the variance in the
solution and, thus, for one factor is relevant for the analysis.
Table 3: Explanation of total variance Factor Eigenvalue Difference Proportion Cumulative
Factor1 1.98719 1.44147 0.8277 0.8277
Factor2 0.54572 0.15173 0.2273 1.055
Factor3 0.39399 0.41488 0.1641 1.2191
Factor4 -0.02089 0.1947 -0.0087 1.2104
Factor5 -0.21559 0.07404 -0.0898 1.1206
Factor6 -0.28963 . -0.1206 1
Table 4 presents the factor loadings (pattern matrix) as per uniqueness. Uniqueness is the
variance that is ‘unique’ to the variable and not shared with other variables. The higher values of
uniqueness for some of the variables imply that these variables are not well explained by the
factors. For example, 81.8% of the variance in ‘total water supply’ is not shared with other
variables in the overall factor model. On the contrary ‘total number of electricity connections’
has low variance not accounted by other variables (43.34%). The values of the factor loadings
9
for all the variables are greater than 0.3. Therefore, we can infer that factor1 is defined by all the
six variables which are then considered to create the infrastructure index3. It is also important to
note that as we are using one factor only, factor rotation which helps to see the underlying
dimensions (scales) more clearly is not suitable as there’s nothing to rotate.
Table 4: Factor loadings (pattern matrix) and unique variances for one factor model
Variable Factor1 Uniqueness
Total road length 0.6481 0.5799
Total number of latrines 0.5841 0.6589
total water supply 0.4266 0.818
Total number of electricity connections 0.7527 0.4334
Total number of hospitals 0.4627 0.7859
Total number of schools, colleges, and universities 0.5131 0.7367
Regression analysis has been carried out to investigate the determinants of city size. Ordinary
Least Square (OLS) as base run and Tobit estimate both have been carried out. Since million-
plus cities form only a small part of the distribution pertaining to all types of cities and towns, a
Tobit function deems appropriate.
Table 5 presents the results of the multiple regression analysis. Models 1 and 2 consider the
logarithm of the city population as the dependent variable. While model 1 gives the OLS
regression results, model 2 presents the Tobit estimates. In model 1, the statistically significant
value of F-statistics indicates that there is a significant relationship between the dependent and
the independent variables. The regression explains 58% of the total variation in the dependent
variable. On the other hand, the likelihood ratio (chi-square of 40.26 with 7 degrees of freedom
and significant at 1 % level indicates that our model as a whole performs well statistically in
comparison to an empty model (i.e., a model with no predictors). Table 5 shows that the OLS
and Tobit regression results are consistent and provide similar results. The results are indicative
that higher average monthly per capita consumption expenditure, total municipal solid waste
generation, and salaried employment have a positive and statistically significant effect on the log
of the city population. Regression model 1 shows that a 1 % increase in average monthly per
capita expenditure increases the city population by 0.22 %. Regression model 2 shows that for
one unit increase in total municipal solid waste generation, there is a 1.23 point increase in the
predicted size of the city population. A one-unit increase in salaried employment is associated
3 Quite importantly, factor1 is mostly related to the city-wise number of electricity connections.
10
with a 0.03 unit increase in the predicted value of the size of the city population. On the
contrary, total registered motor vehicles have a negative effect on the size of the city population.
Table 5: Determinants of million plus city populations
Dependent variable:
Log of city population
VARIABLES OLS Tobit
Model 1 Model 2
Average monthly per capita consumption
expenditure
0.220* 0.220*
(0.130) (0.118)
Total municipal solid waste generation 1.233* 1.233*
(0.683) (0.622)
Infrastructure index -0.103 -0.103
(0.112) (0.102)
Salaried employed persons 0.0295*** 0.0295***
(0.00978) (0.00891)
Self-employed persons 0.0132 0.0132
(0.00984) (0.00897)
Total registered motor vehicles -0.1053** -0.1053***
(0.042) (0.0382)
Incidence of total cognizable crimes -0.0473 -0.0473
(0.0360) (0.0328)
Constant 12.42*** 12.42***
(0.733) (0.668)
F statistics 7.55***
R2 / Pseudo R2 0.575 0.3705
Adj R-squared 0.4992
Log likelihood -34.2032
Likelihood ratio chi2(7) 40.26***
/sigma 0.50096
Observations 47 47
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
The coefficient of -0.105 in regression model 1 indicates that a 10% increase in total registered
motor vehicle decreases a million-plus city population by 1 %. This indicates that congestion of
cities which is measured by total registered motor vehicles deter the size of million-plus cities.
This signifies that million-plus cities are overcrowded in India. However, the city-wise crime
rate and index of infrastructure do not turn out to be statistically significant. This specifies that
million-plus cities have several negative externalities which may not show large variations
across. Similarly infrastructure provision might have been already very high and thus, may not
11
contribute to further population growth as it may be approaching the saturation point. To
generate a new equilibrium with a higher population level massive investment may be required.
Moreover, after the saturation point is reached it may not be possible to expand the city further.
Rather out-migration would unfold a negative relation between infrastructure investment and
population size.
4.1 Cluster analysis
Cluster analysis has been conducted to find out if within the category of million-plus cities
different groups are traceable. Cluster analysis helps us to group similar observations into a
number of clusters based on the observed values of several variables for each individual. In other
words, it is done to identify the set of objects with similar characteristics
We use a hierarchical cluster which is the most common method in cluster analysis. It creates a
series of models with cluster solutions from 1 (all cases in one cluster) to n (each case is an
individual cluster). It is also very useful when the number of observations is less than 100. We
follow the agglomerative clustering in which most hierarchical methods fall into. For this
process, we need to specify a linkage algorithm to define the distance from a newly formed
cluster to other clusters in the solution. For this purpose, we use most commonly used Ward's
minimum variance method. The method combines those objects whose merger increases the
overall cluster variance (i.e., the homogeneity of clusters) to the smallest possible degree. The
approach is usually used in combination with (squared) Euclidean distances. The squared
Euclidean distance increases the importance of a large distance while fading the importance of
small distances.
It is very much important to select clustering variables. To represent a million-plus city structure
we choose several important economic factors. We choose rural to urban migration, wage
structure which is defined by average monthly per capita consumption expenditure, poverty rate,
solid waste generation, number of road accidents, total cognizable crime, infrastructure
availability, employment status, and size of city population to define a million-plus city structure.
12
Table 6: Correlation matrix of clustering variables
migr Mpce phr swm rc tcc self_m self_f casual_m casual_f infras pop
migr 1.00
mpce -0.07 1.00
phr 0.06 -0.74 1.00
swm -0.39 0.06 -0.16 1.00
rc 0.10 0.08 0.04 -0.13 1.00
tcc 0.10 0.23 0.05 -0.22 0.49 1.00
self_m -0.11 -0.48 0.45 0.04 -0.11 -0.08 1.00
self_f -0.13 -0.23 0.20 -0.17 -0.19 -0.15 0.60 0.01
casual_m 0.34 -0.21 0.22 -0.41 0.01 0.11 -0.25 -0.11 1.00
casual_f 0.23 -0.33 0.28 -0.24 0.05 0.14 -0.10 -0.31 0.73 1.00
infras 0.02 -0.04 0.06 0.42 -0.07 0.08 0.03 -0.13 -0.19 -0.09 1.00
pop -0.37 0.48 -0.44 0.37 -0.14 -0.18 -0.27 -0.17 -0.38 -0.48 -0.08 1.00
Note: See table 9 for variable definitions
Table 6 presents the pair-wise correlation coefficients of the clustering variables. The results
show that collinearity is not at a critical level. The variables casual female employment and
casual male employment show the highest correlation of 0.73, which is clearly lower than 0.90
thresholds. This indicates that we can proceed to the analysis using all twelve clustering
variables.
Table 7: The Variance ratio criterion (VRC) and Duda-Hart indices
Duda/Hart Je(2)/J2(1) index VRC
No. of
clusters
Je(2)/Je(1) pseudo T-squared
Calinski/Harabasz
pseudo-F
1 0.8029 11.05
2 0.7367 12.15 11.05
3 0.7054 9.19 13.08
4 0.6511 5.36 11.91
5 0.6132 8.2 11.82
6 0.6511 4.82 11.87
7 0.5565 3.98 11.77
8 0.5039 2.95 11.42
9 0.5235 6.37 11.33
10 0 . 11.32
Now to decide the number of the cluster we depend on graphical and statistical measures. Table
7 suggests that the largest Duda–Hart Je(2)/Je(1) stopping-rule value is 0.8029, corresponding to
the first group. The smallest pseudo-T-squared value is 2.95 for the eight-group solution, but the
13
pseudo-T-squared value for the three-group solution is also low, with a value of 9.19. The three-
group solution with a Calinski–Harabasz pseudo - F value of 13.08 is largest, indicating that the
three-group solution is most distinct compared to any other group solution. The results are also
confirmed by the dendrogram presented in Appendix figure 2.
Table 8: Number of clusters
cluster Frequency Percent Cumulative
1 24 51.06 51.06
2 12 25.53 76.6
3 11 23.4 100
Total 47 100
The output in Table 8 shows that the cluster analysis assigned to all 47 million-plus population
size cities unravels the three segments. There is no city having missing values. The first
cluster comprises 24 cities (51 %), the second cluster 12 cities (26%), and the third
cluster 11 cities (23 %).
Table 9: Comparison of means
cluster Migr mpce phr swm rc tcc self_m self_f casual_m casual_f infras pop
1 24.56 3.24 12.56 0.28 0.73 3.45 33.98 36.65 14.09 13.89 0.02 14.79
2 27.88 2.24 38.50 0.23 0.89 3.72 37.26 36.99 25.82 26.77 0.07 14.17
3 14.81 2.43 27.40 0.29 0.68 3.34 54.10 56.80 8.78 10.51 -0.12 14.58
Total 23.13 2.80 22.65 0.27 0.76 3.49 39.53 41.45 15.84 16.39 0.00 14.58
Note: See table 9 for variable definitions
The mean values for the three clusters are given in Table 9. Comparing the variable means across
the three clusters, we find that among the different variables the first cluster stresses on rural to
urban migration, self-employment male and self-employment female, while the other variables
are less important. The main variables in the second cluster are rural to urban migration, poverty
head count ratio, self-employment male, self-employment female, casual employment male, and
casual employment female. The third cluster considers poverty headcount ratio, self-employment
male, self-employment female, city population size as significant variables. This indicates that
rural to urban migration, poverty rate, employment status, and city population size play the most
important role in clustering the cities into three segments. Finally, Table 10 presents the name of
the million-plus cities that belong to three different clusters.
14
Table 10: Cities divided by different clusters
Cluster City name Average population
size (in millions)
1
Bangalore, Chandigarh, Chennai, Coimbatore, Delhi, Ghaziabad, Hyderabad,
Jodhpur, Kochi, Kollam, Kozhikode, Madurai, Malappuram, Mumbai,
Nagpur, Nashik, Pune, Rajkot, Surat, Thiruvananthapuram, Thrissur,
Tiruchirappalli, Vadodara, Vijayawada
4
2 Agra, Allahabad, Asansol, Aurangabad, Bhopal, Dhanbad, Durg and Bhilai
Nagar, Jabalpur, Jamshedpur, Kannur, Lucknow, Raipur 1.5
3 Ahmedabad, Amritsar, Gwalior, Indore, Kanpur, Kolkata, Meerut, Patna,
Ranchi, Srinagar, Varanasi 3
5. City, Migration and Poverty
Various economic theories suggest that structural change which is an inevitable component of
economic growth not only involves shift in the value added and work force composition but also
the transfer of population from the rural to the urban areas. In the Lewis (1954) model, shifting
of labor from agriculture sector to manufacturing triggers growth by generating an investible
surplus; and in the process of structural change it results in both internal and external
agglomeration economies. The Todaro (1969) model explained rural-urban migration as a
function of income differential adjusted for the probability of finding a job. The Harris-Todaro
(1970) model attempted to explain the phenomenon of accelerating rural-urban labor migration
despite the existence of positive marginal products in agriculture and significant levels of urban
unemployment.
The intrastate migration rate is much higher in magnitude than the interstate migration rate in
India (Mitra and Murayama, 2009) Male and female migration rates are closely interconnected,
irrespective of whether they migrate from the rural areas within the states or other states though
the social and cultural diversity in India stands as a major hindrance to population mobility. The
migration rate in the urban areas and the urbanization level are positively associated, very
moderately though. But with an increase in the city size, the migration rate rises mainly because
employment prospects are better in large cities due to agglomeration effects (Mitra, 2019).
Despite being a source of employment, MGNREGA has not been able to check the out-migration
from the developed region because of higher market wage rates in urban destination (Ahuja et al.
2011). Besides, an increasing level of education of the migrants acts as the main pull factor
(Sridhar et al., 2013). The overall rural to urban migration rate is low due to the lowering of the
15
gap between urban–rural health, infrastructure, employment opportunities, and economic
conditions (Tripathi, forthcoming). Also, as Chauvin et al. (2017) argued, India’s rural-urban
migration is limited by strong place-based preferences such as those related to caste-based social
networks. Hence, if the contact person is located in a large city the potential migrant will have a
higher probability of moving to the same city.
Is this migration process beneficial? India: Urban Poverty Report 2009 (Government of India,
2009) indicated that the relationship between poverty and migration is not established clearly and
it is evidenced that middle and higher-income groups have a higher propensity to move. Based
on better education and skills, the most effective group of migrants is the urban to urban stream
directed to large and metropolitan cities. Cali and Menon (2013) noted the positive spillovers of
urbanization on the rural economy rather than the movement of the rural poor to the urban areas.
This rural poverty-reducing effect of urbanization is primarily explained by the increased
demand for local agricultural products.
While large cities may offer better employment opportunities compared to the small towns even
within the informal sector, excess supplies of labour and the segmentation of the labour market
due to information asymmetry may restrict the pace of upward mobility. Thus, large cities may
still comprise poverty. It would be therefore worthwhile to pursue further analysis in this regard.
In order to study the relationship among migration, poverty and city size factor analysis has been
carried out. We evaluate and ensure the validity of the data by using the Kaiser-Meyer-Olkin
(KMO) index and the Bartlett’s test of Sphericity. The KMO value is 0.569, indicating that factor
analysis is relatively suitable. The significance probability of Bartlett’s test of Sphericity is 0.000
< 0.01 which indicates that the parameters are not related, and thus, are suitable for factor
analysis. Table 11 presents the test results.
Table 11: KMO and Bartlett’s test
KMO measure of sampling adequacy 0.569
Bartlett’s test of sphericity Approximate chi-square 217.756
df 66
Sig. 0.000
KMO, Kaiser-Meyer-Olkin.
16
Table 12 presents the eigenvalues, the percentage of variance, and the cumulative percentage of
variance associated with each other. It reveals that the first three factors explain
approximately 88.53% of the total variance, and hence, consideration of three factors is
appropriate for the analysis. Table 7 and appendix figure 3 also suggest that we should retain
three factors as the Eigenvalues are equal or greater than 1.
Table 12: Explanation of total variance Factor Eigenvalue Difference Proportion Cumulative
Factor1 2.76788 0.72829 0.4244 0.4244
Factor2 2.03959 1.07374 0.3128 0.7372
Factor3 0.96585 0.09586 0.1481 0.8853
Factor4 0.86999 0.49338 0.1334 1.0187
Factor5 0.37661 0.14052 0.0578 1.0765
Factor6 0.23609 0.15491 0.0362 1.1127
Factor7 0.08118 0.1146 0.0124 1.1251
Factor8 -0.03343 0.05349 -0.0051 1.12
Factor9 -0.08691 0.09845 -0.0133 1.1067
Factor10 -0.18536 0.0397 -0.0284 1.0782
Factor11 -0.22506 0.06017 -0.0345 1.0437
Factor12 -0.28522 . -0.0437 1
Table 13 presents the estimated values of factor score coefficients and the rotated orthogonal
varimax, respectively. A varimax rotation attempts to maximize the squared loadings of the
columns. The higher values of uniqueness for infrastructure index, the total number of road
accidents, and crime rate indicate that these variables are not well explained by the factors. On
the other hand, lower values of average monthly per capita expenditure, casual female
employment, self-employment male, and self-employment female show that these variables are
well explained by the factors under consideration.
Table 13 suggests that migration and city population size are moderately associated. However,
there is a negative association between the two, indicating that migration tends to decline with a
rise in city population size. Large cities also reveal a lower incidence of poverty and migration
and city poverty are positively associated, very mildly though. Large cities also show a lower
level of self-employment for males, self-employment for females, casual employment for males,
and casual employment for females in factors 1 and 2, respectively. The relationships are
moderately associated.
17
Table 13: Loading for varimax rotated factor matrix of three-factor model
Variable Factor1 Factor2 Factor3 Uniqueness
Total migration from rural India to million plus cities
(migr) -0.0133 0.3751 0.295 0.7721
Average monthly per capita consumption expenditure
(mpce) -0.7948 -0.3097 0.1224 0.2574
City-wise poverty headcount ratio (phr) 0.7059 0.2955 0.0369 0.413
Total municipal solid waste generation (swm) -0.0743 -0.2551 -0.7016 0.4372
Number of road accident (rc) -0.1696 0.1297 0.3008 0.8639
Incidence of total cognizable crimes (tcc) -0.2013 0.1766 0.3669 0.7937
Self-employed male (self_m) 0.7431 -0.2933 0.0542 0.3588
Self-employed female (self_f) 0.5602 -0.4497 0.3015 0.393
Casual employed male (casual_m) 0.0377 0.7423 0.2688 0.3753
Casual employed female (casual_f) 0.1183 0.8398 0.052 0.278
Infrastructure index (infras) 0.0422 0.0177 -0.4133 0.8271
Log of city population (pop) -0.4706 -0.4716 -0.3145 0.4572
Finally, Table 14 presents the factor rotation matrix. The conversion matrix estimates the rotated
factor loadings (RFL): RFL = Factor loadings* Factor rotation. This matrix produces rotated
factor matrix. The result indicates larger rotations as off-diagonal elements are larger.
Table 14: Factor rotation matrix
Factor1 Factor2 Factor3
Factor1 0.6118 0.7026 0.3634
Factor2 -0.7896 0.57 0.2272
Factor3 0.0475 0.4259 -0.9035
6. Conclusion and policy implications
City sizes play a pivotal role in the context of urbanization, as India’s urban population is mainly
concentrated in and around large cities. In this paper on million plus cities, we estimate the city
size distribution by applying Pareto distribution framework. Though the rank-size rule does not
apply to the million-plus cities, there is evidence to suggest somewhat concentration of economic
activities rather than a highly diversified system. In fact, the cluster analysis also divides cities
into only three categories rather than suggesting that each one is different from the other. Cities
18
can be grouped into three clusters with the average population sizes of 4, 3 and 1.5 million
respectively.
We consider infrastructure, city structure, employment type, wage rate and diseconomies such as
crime rate as the major determinants of city size. City structure envisaged in terms of waste
generation and employment status, wage rate captured in terms of consumption expenditure, and
capacity measured in terms of registered motor vehicles turn out to be the important
determinants. Crime and index of infrastructure are not statistically significant, though.
As regards the impact of million-plus cities on migration and poverty, we noted that migration
tends to decline with a rise in city size. However, large cities reveal a lower incidence of poverty,
which is suggestive of the beneficial effects of agglomeration economies. Though migration
tends to raise the incidence of poverty at the place of destination, positive externalities associated
with city size contribute to livelihood creation and poverty reduction. These findings have
important policy implications. Facilitating the rural migrants in cities in terms of basic amenities
and land tenure can contribute to poverty reduction in a more cost effective manner in
comparison to the anti-poverty measures implemented at the place of rural origin.
Appendix
Measurement of variables, data sources, and description of data
City population: Million-plus city population data is collected from Census of India, 2011.
Website: https://www.census2011.co.in/urbanagglomeration.php
Employment: National Sample Survey does not provide city level employment data. Urban
samples of a city district (i.e. the district to which the sample city is located) are considered to
measure the distribution of employment. The percentage distribution of all usual status (principal
status+ subsidiary status) of age 15 years and above in different employment categories are
estimated. Source: Unit level data of NSS 68th Round on Employment and Unemployment in
2011-12.
Total road length: Both Kachcha road length and Pucca road length are considered for the
measurement of total road length of a city. Source: Town amenities, District Census Hand Book,
Census of India 2011. Website: http://censusindia.gov.in/2011census/dchb/DCHB.html
19
Electricity connection: Total number of electricity connections in domestic, industrial,
commercial, road lighting, electricity, and other connections. Source: Town amenities, District
Census Hand Book, Census of India 2011.
Number of Latrines: Total number of pit, flush/pour, services, and other latrines. Source: Town
amenities, District Census Hand Book, Census of India 2011.
Total water supply: Total protected water supply in city. Source: Town amenities, District
Census Hand Book, Census of India 2011.
Total hospitals: It includes allopathic hospitals, alternative medicine hospitals, dispensary/health
Centers, family welfare centers, maternity and child welfare centers, maternity homes, TB
hospitals/ clinic, and nursing homes Source: Town amenities, District Census Hand Book,
Census of India 2011.
Total number of schools, colleges, and universities: It includes all the private and governments’
school, colleges and universities of a city. Source: Town amenities, District Census Hand Book,
Census of India 2011.
Average monthly per capita consumption expenditure (MPCE): The MPCE is measured by the
modified mixed reference period (MMRP). MPCE is estimated using urban samples of a city
district. Source: Author’s calculation using unit level data of the NSS 68th Round on
consumption expenditure of 2011-12.
Poverty headcount ratio: Head count ratio (HCR) is the proportion of a population that exists, or
lives, below the poverty line. For measuring city poverty rate, we use the Rangarajan committee
–recommended poverty line in 2011-12 by considering monthly per capita consumption
expenditure based on MMRP. However, as India’s official estimates do not provide the city-level
poverty line, state-specific urban poverty lines have been used for measuring district-level
(which is used as proxy of a city) urban poverty for the districts located in the corresponding
states. Source: Author’s calculation using unit level data of the NSS 68th Round on consumption
expenditure of 2011-12.
Migration: City-specific percentage of migration is defined by the total number of migrants from
India’s rural areas to a particular city district with duration of residency from less than one year
to more than 10 years divided by the total population of that city. Migration Table, Census of
India 2011, The Government of India.
Registered motor vehicles: Data on city wise registered motor vehicles in 2012 is collected
from Statistical Year Book India 2017, Ministry of Statistics and Programme Implementations,
The Government of India. Website: http://mospi.nic.in/statistical-year-book-india/2017/189.
Cognizable crimes: Data on city wise incidence of total cognizable crimes in 2011 is collected
from National Crime Records Bureau, Ministry of Home Affairs, The Government of India.
20
Solid Waste Generation: Biodegradable and Non-biodegradable waste are considered. Total
municipal solid waste generation per day in 2011 is collected from Central Pollution Control
Board (CPCB), the Ministry of Environment, Forest and Climate Change., The Government of
India.
Road Accidents: Data is collected from the Ministry of Road Transport and Highways, The
Government of India.
Appendix Table 1: Description of data
Variable Mean
Standard
Deviation Minimum Maximum
Coefficient of
variation
Log of city population 14.58 0.78 13.84 16.73 5.33
Total road length per 1000 population
(kilometers) 0.85 0.58 0.03 2.03 68.18
Total number of latrines per 1000 population 121.86 81.76 0.89 414.87 67.09
Total water supply per 1000 population (in
kiloliters) 99.47 122.86 0.46 636.23 123.51
Total number of electricity connections per
1000 population 187.92 75.13 16.05 328.46 39.98
Total number of hospitals per 1000 population 0.17 0.25 0.00 1.36 145.63
Total number of schools, colleges, and
universities per 1000 population 0.56 0.40 0.01 2.48 71.83
Average monthly per capita consumption
expenditure (in thousands rupees) 2.80 0.82 1.44 4.89 29.29
Total municipal solid waste generation per
1000 population (tons per day) 0.27 0.15 0.01 0.59 53.98
Salaried employed persons (%) 43.38 13.71 12.61 67.85 31.60
Total registered motor vehicles per 1000
population 418.21 217.39 35.34 1139.60 51.98
Incidence of total cognizable crimes per 1000
population 3.49 2.48 1.22 16.35 70.91
Migration from rural India to a million plus
cities (%) 23.13 13.96 2.32 62.51 60.36
Poverty headcount ratio (%) 22.65 14.22 1.20 72.54 62.78
Number of road accidents per 1000 population 0.76 0.57 0.05 2.30 74.62
Self-employed male (%) 39.53 11.55 21.44 80.95 29.23
Self-employed female (%) 41.45 13.53 20.02 85.90 32.63
Casual employed male (%) 15.84 11.76 1.14 49.22 74.27
Casual employed female (%) 16.39 11.33 1.02 51.99 69.15
Note: Calculation is based on 47 observations
21
Appendix Figure 1. Scree plot of Eigen values after factor solution
Appendix Figure 2: Dendrogram for wards linkage cluster analysis
Appendix Figure 3: Scree plot of Eigen values after factor solution
-.5
0.5
11.
52
Eig
enva
lues
1 2 3 4 5 6Number
0
5000
1000
015
000
2000
0
L2sq
uare
d dis
simila
rity m
easu
re
G1n=9
G2n=7
G3n=8
G4n=3
G5n=2
G6n=6
G7n=1
G8n=4
G9n=5
G10n=2
-10
12
3
Eig
enva
lues
0 5 10 15Number
22
References
Ahuja, U. R.; Tyagi, D.; Chauhan, S.; Chaudhary, K. R. Impact of MGNREGA on rural
employment and migration: A study in agriculturally-backward and agriculturally-
advanced districts of Haryana. Agricultural Economics Research Review 2011, 24
(Conference Number), 495-502.
Bhalla, S. S.; Kaur, R. Labour force participation of women in India: Some facts, some Queries;
Asia Research Centre Working Paper 40, London School of Economics & Political
Science, London, 2011.
Bhanumurthy, N.R.; Mitra, A. Declining poverty in India: A decomposition analysis. Indian
Journal of Labour Economics 2004, 47, 311-21.
Calem, P.S.; G.A. Carlino, G.A. Urban agglomeration economies in the presence of technical
change. Journal of Urban Economics1991, 29, 82-95.
Calì , M.; Carlo, M. Does urbanization affect rural poverty? Evidence from Indian districts. The
World Bank Economic Review 2013, 27, 171-201.
Chen, M. A.; Raveendran, G. Urban employment in India: Recent trends and patterns. Margin:
The Journal of Applied Economic Research 2012, 6, 159-179.
Chauvin, J. P.; Glaeser, E.; Ma, Y.; Tobio, K. What is different about urbanization in rich and
poor countries? Cities in Brazil, China, India and the United States. Journal of Urban
Economics 2017, 98, 17-49.
Echeverri-Carroll, E.L.; Ayala, S.G. Urban wages : Does city size matter? Urban Studies 2011,
48, 253–271.
Federation of Indian Chambers of Commerce & Industry. 2011. Urban infrastructure in India.
FICCI,New Delhi, India, 2011.
Gangopadhyay, S.; Lanjouw, P.; Vishwanath, T.; Yoshida, N. Identifying pockets of poverty:
Insights from poverty mapping experiments in Andhra Pradesh, Orissa and West Bengal.
Indian Journal of Human Development 2010, 4, 5-28.
Government of India. India: Urban Poverty Report 2009, Ministry of Housing and Urban
Poverty Alleviation and Oxford University Press, New Delhi, 2009.
Harris, J.R.; Todaro, M.P. Migration, Unemployment and Development: A Two-sector Analysis.
American Economic Review 1970, 60,126–142.
Haynes, R. M . Crime rates and city size in America. Area 1973, 5, 162-165.
Infrastructure Development Finance Company. India Infrastructure Report 2011. Water: Policy
and Performance for Sustainable Development. Oxford University Press, New Delhi, India,
2011. Available online: http://www.idfc.com/pdf/report/IIR-2011.pdf (accessed on 19
November 2013)
Institute for Human Development (IHD). India labour and employment report 2014: Workers in
the era of globalization, Academic Foundation & Institute for Human Development, 2014.
Indian Institute for Human Settlements (IIHS). “Urban India 2011: Evidence”, Bangalore, India,
2012, accessed on 5 June 2013 from iihs.co.in/wp-content/uploads/2013/12/IUC-Book.pdf.
Krugman, P. 1991). Increasing returns and economic geography. Journal of Political Economy
1991, 99, 483-499.
Lewis, W. A. Economic development with unlimited supplies of labor. The Manchester School
of Economic and Social Studies 1954, 22, 139-191.
Maiti, M. (2015): “Understanding the employment challenges in India”, International Research
Journal of Social Sciences 2015, 4, 1-8.
23
Mathur, O.P. Impact of globalization on cities and city‐related policies in India. In
Globalization and Urban Development, Richardson, H.W., Bae, C.‐H.C., Eds.; Springer:
Berlin, Grmany, 2005, pp. 43–58.
Mathur, O.P. City-Size Distributions in a Quasi-Open Economy: The India Evidence. In Cities of
Dragons and Elephants: Urbanization and Urban Development in China and India, Wan,
G., Lu, M., Eds.; Oxford University Press, 2019, pp. 103-129.
Mazumdar, D.; Son, H.H. Vulnerable groups and the labour market in Thailand: Impact of the
Asian financial crisis in the light of Thailand’s growth process, paper presented at a
workshop on ‘Impact of globalisation on the labour markets’ at NCAER, Delhi, 2002.
Mehrotra, S.; Parida, J; Sinha, S.; Gandhi, A. Explaining employment trends in the Indian
economy: 1993-94 to 2011-12. Economic and Political Weekly 2014, 49, 49-57.
Mera, K. Regional production functions and social overhead capital: An analysis of the Japanese
case. Regional and Urban Economics 1973, 3,157-186.
Mitra, A. Rural to urban migration and urban labour market: the case of India. In Cities of
Dragons and Elephants: Urbanization and Urban Development in China and India, Wan,
G., Lu, M., Eds.; Oxford University Press, 2019, pp. 175-218.
Mitra, A.; Murayama, M. Rural to urban migration: A district level Analysis for India.
International Journal of Migration, Health and Social Care 2009, 5, 35-52.
Mitra, A.; Singh, J. Rising Unemployment in India: A Statewise Analysis from 1993–94 to
2017–18, Economic and Political Weekly 2019, 54.
Moomaw, R.L. Is population scale a worthless surrogate for business agglomeration economies?
Regional Science and Urban Economics 1983,13, 525-545.
Moscona, J. The impact of India’s green revolution: An empirical investigation of modern
agricultural development, Unpublished Manuscript, 2017.
Nair, V.D. Infrastructure development in India, Dr. Marri Channa Reddy Human Resource
Development Institute, Hyderabad, 2012.
O’Clery, N.; Lora, E. City Size, Distance and Formal Employment, CID Research Fellow and
Graduate Student Working Paper No. 77, October 2016.
Pandey, S.M. Nature and determinants of urbanization in a developing economy: The case of
India. Economic Development and Cultural Change 1977, 25, 265-278.
Papola, T. S.; Sahu, P. P. (2012): Growth and structure of employment in India long-term and
post reform performance and the emerging challenge”, Institute for Studies in Industrial
Development, New Delhi, 2012. Available online:
http://isidev.nic.in/pdf/ICSSR_TSP_PPS.pdf (accessed on 5 May 2014).
Pradhan, R. 2007. Does infrastructure play role in urbanization: Evidence from India. Indian
Journal of Economics and Business 2007, 6, 81–92.
Ramaswamy, K. V.; Agrawal, T. Services-led growth, employment, skill, and job quality: A
study of manufacturing and service sectors in urban India. In M. Dev (eds). India
Development Report 2012–13, New Delhi, Oxford University Press: 116-131, 2012.
Seitz, H. The impact of the provision of urban infrastructures on the manufacturing industry in
cities, paper presented at the 33rd European Congress of the Regional Science Association,
Moscow, Russia, 1993.
Sridhar, K.S. Determinants of city growth and output in India. Review of Urban and Regional
Development Studies 2010, 22, 22–38
Sridhar, K. S.; Reddy, V.; Srinath, P. Is it push or pull? Recent evidence from migration into
Bangalore, India, International Migration and Integration 2013, 14, 287–306.
24
Tiebout, C.M. A pure theory of local expenditures. The Journal of Political Economy 1956, 64,
416–424.
Todaro, M. P. A model of labor migration and urban unemployment in less developed countries.
American Economic Review 1969, 59, 138–148.
Toutain, O.; Gopiprasad, S. 2006. India Infrastructure Report 2006. Oxford University Press,
New Delhi, 2006. Available online: http://www.idfc.com/pdf/report/IIR-2006.pdf (accessed
on 10 November, 2012).
Tripathi, S. Do large agglomerations lead to economic growth? Evidence from urban India.
Review of Urban and Regional Development Studies 2013a, 25, 176-200.
Tripathi, S. Does higher economic growth reduce poverty and increase inequality? Evidence
from urban India. Indian Journal of Human Development 2013b, 7, 109-137.
Tripathi, S. Does a Higher Level of Infrastructure Increase Population in Large Agglomerations?
Evidence from India, Review of Urban and Regional Development Studies 2018, 30, 145-
168.
Tripathi, S. (forthcoming). Why is the Rural to Urban Migration Rate in India so Low? An
Empirical Analysis, Review of Regional Studies.
World Bank. Perspectives on Poverty in India: Stylized Facts from Survey Data”, India Poverty
Assessment, Poverty Reduction and Economic Management Network, The World Bank,
Washington DC, 2010.