examining the income convergence among indian...

Income Convergence among Indian States 1

Examining the Income Convergence among IndianStates: Time Series Evidence with Structural

Breaks

Ankita MishraSchool of Economics, Finance and Marketing, RMIT University

ankita.mishra@rmit.edu.au

Vinod Mishra*Department of Economics, Monash University

vinod.mishra@monash.edu

10th Annual Conference on Economic Growth and DevelopmentIndian Statistical Institute - New Delhi

December 18-20, 2014

Income Convergence among Indian States 2

Outline

1 Introduction

2 Existing Literature

3 Motivation

4 Data

5 Econometric Methodology and Results

6 Discussion

Income Convergence among Indian States Introduction 3

Introduction

New Growth Theory envisioned the poor countries/regionscatching up’ to rich countries/regions in levels of per capitaIncome.

Pay offs from using additional capital or better technology aremore for a poor economy, hence poor economy will grow at afaster rate compared to a rich economy.

.... and (eventually) able to catch-up i.e. per capita incomeswill converge.

Income Convergence among Indian States Existing Literature 4

Empirical Studies

Has this catch-up or convergence actually occurred ?

For different groups of countries: Mankiw et. al (1992), Evans(1996, 1997)For diverse regions/states within a single large country(primarily United States): Young et. al (2008), Carlino andMills (1993)

The two main techniques, to investigate the convergencehypothesis:

Cross-Sectional growth equation estimations: Barro andSala-i-Martin (1992, 1995), Mankiw et. al (1992))Time-Series unit roots testing: Bernard and Durlauf (1995),Carlino and Mills (1993), Fleissig and Strauss (2001) andStrazicich et al. (2004)

Income Convergence among Indian States Existing Literature 5

The Indian Context

The notion of convergence defined as inclusive growth holdspivotal place in Indian central planning.

Notwithstanding with objectives of central planning, regionalimbalance in growth is quite acute for India.

Bandyopadhyay (2011): ... The richest states in India, forexample, Gujarat and Maharashtra are akin to middle incomecountries like Poland and Brazil in their level of developmentwhile the poorest states of Bihar, Uttar Pradesh and Orissaare more analogous to that of some of the poorestSub-Saharan African countries.

Income Convergence among Indian States Existing Literature 6

The Indian Context

Majority of the studies on regional economic growth in Indiafound little support for absolute convergence.

Ghosh et. al (2013): Increasing divergence in regional percapita income.

Found little support for Club convergence hypothesis forIndian states: Baddeley et al. (2006), Bandyopadhyay (2011)and Ghosh et. al (2013).

Cross section growth convergence equation approach: Bajpaiand Sachs (1996), Cashin and Sahay (1996), Nagaraj et al.(2000), Aiyar (2001) and Trivedi (2003).

Stochastic Convergence: Ghosh et al (2013), Bandyopadhyay(2011)

Stochastic kernel density techniques in Bandyopadhyay (2011)Nonlinear transition factor model in Ghosh et al (2013)

Income Convergence among Indian States Motivation 7

Contribution of this Study

Bandyopadhyay (2011): ... time series approaches along thelines of Carlino and Mills (1993) that estimate the univariatedynamics of income remain incomplete in describing thedynamics of the entire cross-section

Ghosh (2013): unit root tests employed in stochasticconvergence literature suffer from low power due to ignoringthe possibility of the presence of structural breaks in thecontext of a single time series or in panel data framework.

Addressing the concerns raised in above studies, using thelatest advancements in time series approach to examine theincome convergence hypothesis among 17 Indian states for theperiod 1960-2011.

This paper finds evidence in support of income convergenceamong Indian states, contrary to what earlier studies in theliterature examining regional income convergence in Indiafound.

Income Convergence among Indian States Motivation 8

Contribution of this Study

Testing methodology is not prone to rejections of the null inthe presence of a unit root with break(s).

Rejection of null hypothesis (of unit root) unambiguouslyimplies stationarity.

Panel versions of unit root tests with structural breaks thatexploit both the cross-sectional and time series informationavailable in the data.

In a situation where univariate unit root tests (with or withoutstructural breaks) give conflicting results, overall incomeconvergence can still be ascertained.

The issue of cross-sectional dependence.

Income Convergence among Indian States Data 9

Data Description

Per capita Net State Domestic Product (NSDP) forseventeen major states of India for the period 1960-61 to2011-12.

Source: Indiastat database.

All series expressed in INR and converted to the common baseperiod of 2004-05.

Av Per Capita NSDP States

Below Rs. 10,000 Bihar, UP, Assam, MP, Orissa, Manipur,Rajasthan, Tripura and WB

Rs. 10,000 - 15,000 Gujarat, Tamil Nadu, Kerala, Karnatakaand Andhra Pradesh

Above Rs. 15,000 Haryana, Maharashtra and Punjab

Income Convergence among Indian States Data 10

Descriptive Statistics

Income Convergence among Indian States Data 11

Variable construction

ln(Relative Per Capita NSDPit = Per Capita NSDPitAverage Per Capita NSDPt

)

Unit Root −→ Shocks to income of state i relative to theaverage income will be permanent. Hence, per capita incomeof state i will diverge from the average per capita income ofthe group.

Stationarity or Mean reversion −→ Shocks to income of statei relative to the average income will be transient and over thelong run, per capita income of state i will converge to theaverage per capita income of the group.

Income Convergence among Indian States Econometric Methodology and Results 12

Cross-Sectional Dependence

This transformation has the advantage that it removes thecross-sectional shocks that affect all the states in the panel.Therefore, any structural breaks identified in the transformedseries would be state specific (Mishra and Smyth (2014)).

Pesaran (2004) cross-sectional dependence (CD) test on theper capita NSDP series is conducted, before and aftertransforming it into relative per capita NSDP.

The null hypothesis in CD test is that the series arecross-sectionally independent.

Income Convergence among Indian States Econometric Methodology and Results 13

Pesaran (2004) CD Test

Income Convergence among Indian States Econometric Methodology and Results 14

Conventional Unit Root Tests

Augmented Dickey Fuller (ADF) test: Dickey and Fuller,1979.

KPSS stationarity test: Kwiatkowski et al., 1992.

LM unit root test: Schmidt and Phillips, 1992.

The null hypothesis for ADF and LM unit root tests is unitroot, whereas the null of KPSS test is stationarity.

Income Convergence among Indian States Econometric Methodology and Results 15

Conventional Unit Root Tests

Income Convergence among Indian States Econometric Methodology and Results 16

Conventional Unit Root Tests

Augmented Dickey Fuller (ADF) test: The null of unit rootcannot be rejected in any of the transformed series at thetraditional levels of significance, suggesting no evidence ofconvergence in per capita incomes.

KPSS stationarity test: The null of stationarity is rejected for11 out of 17 series.

LM unit root test: Fail to reject null of unit root in 13 out of17 cases.

The overall conclusion that seems evident, based on thesetests results is that, there is no indication of convergence inper capita incomes for Indian states in the sample.

Income Convergence among Indian States Econometric Methodology and Results 17

Conventional Unit Root Tests

Income Convergence among Indian States Econometric Methodology and Results 18

Unit Root Tests with two structural breaks

Lee and Strazicich (2003) LM Unit root

Two endogenous breaks in trend and intercept (Model CC)LM unit root test has the advantage over ADF-typeendogenous break tests (Zivot and Andrews (1992), Lumsdaineand Papell (1997)), that it is unaffected by breaks under thenull of unit root.ADF type endogenous break unit root tests, data series can beconcluded as trend stationary when in reality it isnon-stationary with breaks, giving rise to spurious rejectionproblem (Smyth, Nielsen and Mishra (2009)).

Carrion-i-Silvestre and Sanso (2007) KPSS Unit root test

Two endogenous breaks in trend and intercept (Model CC)The null is stationarity with/without structural breaks.

Income Convergence among Indian States Econometric Methodology and Results 19

Unit Root Tests with two structural breaks

Income Convergence among Indian States Econometric Methodology and Results 20

Unit Root Tests with two structural breaks

Lee and Strazicich (2003) LM Unit root

The null of unit root in relative per capita NSDP series isrejected for 12 states.Comparing these results with Schmidt and Philips LM Unitroot test, it is noted that the number of states for which thenull of unit root can be rejected increases dramatically to 12out of 17 compared to 4 out of 17, when the structural breaksweren’t allowed in the data.

Carrion-i-Silvestre and Sanso (2007) KPSS Unit root test

The KPSS test fails to reject the null hypothesis of stationarityfor 13 states.

Test results point to the considerable, evidence of convergencein per capita incomes of Indian states.

Income Convergence among Indian States Econometric Methodology and Results 21

Unit Root Tests with two structural breaks

Income Convergence among Indian States Econometric Methodology and Results 22

Panel unit root tests with structural breaks

Carrion-i-Silvestre et al (2005) Panel KPSS with MultipleBreaks

This test has the null of stationarity, hence takes into accountthe criticism by Bai and Ng (2004) that for many economicapplications it is more natural to take stationarity as the nullhypothesis, rather than non-stationarity.It allows for structural shifts in the trend of the individual timeseries in the panel and permits each state in the panel to havea different number of breaks at different dates.

Panel LM unit root tests with structural breaks (IM et al(2005))

The null hypothesis of panel unit root.

Income Convergence among Indian States Econometric Methodology and Results 23

Panel unit root tests with structural breaks

Income Convergence among Indian States Discussion 24

Location of Breaks

Except Bihar, Gujarat and Tripura, the first structural break inrelative income occurred during the period of 1966-1979.

Three economic crises during this period, one in 1965-66,second in 1973-74 and third in 1979-80.Wars with Pakistan in 1965 and 1971) and international oilcrisis of 1973 and 1979. , the . This can be associated with thepolitical regime shift in both of these states. Congress partylargely ruled these states since independence. However, in1990 the power shifted from Congress party to Bhartiya JanataParty (BJP) in Gujarat and Janata Dal (JD) in Bihar. Sincethen Congress party never got elected again in these states.

For Bihar and Gujarat:

First structural break occurs in early 1990s.In 1990 the power shifted from Congress party to BhartiyaJanata Party (BJP) in Gujarat and Janata Dal (JD) in Bihar.Since then Congress party never got elected again in thesestates.

Income Convergence among Indian States Discussion 25

Location of Breaks

Except Bihar, Gujarat and Orissa the second structural breakin relative income occurred during the period of 1989-1999.

This period of development is characterised with unsustainablelevel of government spending, resulting in mounting internaland external debt and expenditure on subsidies giving rise toseverest balance of payment crisis in India in 1991Rajiv Gandhi’s assassination in May, 1991.Iraqi invasion of Kuwait in August 1990 and subsequent firstgulf war.Initiation of major structural and economic reforms in 1991.Political uncertainty between 1996 to 1999.

Income Convergence among Indian States Discussion 26

Andhra Pradesh

Income Convergence among Indian States Discussion 27

Assam

Income Convergence among Indian States Discussion 28

Bihar

Income Convergence among Indian States Discussion 29

Gujarat

Income Convergence among Indian States Discussion 30

Haryana

Income Convergence among Indian States Discussion 31

Karnataka

Income Convergence among Indian States Discussion 32

Kerala

Income Convergence among Indian States Discussion 33

Madhya Pradesh

Income Convergence among Indian States Discussion 34

Maharashtra

Income Convergence among Indian States Discussion 35

Manipur

Income Convergence among Indian States Discussion 36

Orissa

Income Convergence among Indian States Discussion 37

Punjab

Income Convergence among Indian States Discussion 38

Rajasthan

Income Convergence among Indian States Discussion 39

Tamil Nadu

Income Convergence among Indian States Discussion 40

Tripura

Income Convergence among Indian States Discussion 41

Uttar Pradesh

Income Convergence among Indian States Discussion 42

West Bengal