Information about Asia and the Pacific Asia y el Pacífico
Chapter

2 Is Economic Growth Leaving Some States Behind?

Author(s):
Catriona Purfield, and Jerald Schiff
Published Date:
August 2006
Share
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details
Author(s)
Catriona Purfield

Despite India’s recent strong economic performance there is a growing concern that growth has benefited India’s richer states, leaving the poorer states lagging further and further behind. As India’s poorest states are also its most populous, one concern is that unless these states begin to share in the benefits of growth, an increasing proportion of the population will be left in poverty and subject to rising inequality, leading to social, political, and economic difficulties. Moreover, as many perceive that globalization and economic liberalization contributed to this state of affairs, economic divergence could erode support for further economic reform and opening of the Indian economy. These concerns gain even greater traction when one considers that about 60 percent of the forecast 620 million increase in the Indian population between now and 2051 is expected to occur in three of its poorest states: Bihar, Madhya Pradesh, and Uttar Pradesh (Visaria and Visaria, 2003).

A rich literature uses state data to test whether growth in regions within India has converged or diverged over time. However, as is often the case in such studies, the results for India are conflicting. For example, Cashin and Sahay (1996) and Aiyar (2001) find evidence of convergence after controlling for differences in initial economic conditions, but Rao, Shand, Kalirajan (1999), Bajpai and Sachs (1996), and Sinha and Sinha (2000) find divergence. Aiyar (2001) also observes that education and investment helped to reduce cross-state income divergence, while Cashin and Sahay (1996) found fiscal transfers were a significant equalizing force. Various studies have also made opposing claims about the impact of the globalization and economic reform post-1991 on income convergence, although few have conducted rigorous statistical tests of this hypothesis. Bhattacharya and Sakthivel (2004) and Kumar (2004) assert that the reforms of the 1990s exacerbated the gap between richer and poorer states, while Ahluwalia (2002) claims that reforms in the 1990s helped reduce the gap. In light of the wide range of evidence, this chapter seeks to shed light on the debate by asking two related questions. First, have poorer states in fact fallen further behind richer states, particularly since the 1990s? Second, why have certain states performed better than others? If state-level economic policies have an impact on growth, better policies could help laggard states catch up. This chapter first presents stylized facts about growth across Indian states and then assesses empirically the question of convergence and the impact of state policy on growth and poverty before drawing conclusions.

Stylized Facts About Growth in India

The disparity in economic conditions across Indian states is large and growing (Table 2.1). Over the past three decades, the ranking of states by income into poor, medium, and rich has changed remarkably little and although poverty has declined, it has become more spatially concentrated. We highlight five key facts about the pattern of development across states.

Table 2.1.Then and Now: Summary Income, Growth, and Poverty Indicators, 1970–2004
Real Per Capita Net State Domestic Product, 2000–04Real Per Capita Income (In rupees)Income Rank inReal Per Capita Income of Richest to Poorest (Ratio)Real Per Capita Income Growth (In percent)Population Growth Rate (In percent)Head Count of Total Poverty (In percent)Overall Literacy Rate (In percent)
(In rupees)2003–041970–741990–941970–742000–041970–20041970–200420001977–7820011973
Poorest states
14 Bihar3,55314141.01.01.91.446.961.647.523.4
13 Uttar Pradesh5,70213121.81.51.42.133.049.157.426.9
12 Orissa6,48711132.11.61.41.746.370.163.632.4
11 Madhya Pradesh8,28412111.82.12.61.336.861.864.128.2
10 Rajasthan8,571992.32.21.52.720.437.461.023.9
Middle-income states
9 Andhra Pradesh11,3331082.22.93.11.818.839.361.129.8
8 West Bengal11,7718102.42.92.91.932.160.569.240.6
7 Kerala12,109472.93.12.51.314.552.290.971.9
6 Karnataka13,141762.43.43.21.925.648.867.038.5
Richest states
5 Tamil Nadu12,976652.83.52.81.421.554.873.547.1
4 Gujarat16,779542.83.93.22.115.641.266.443.7
3 Haryana15,721223.24.02.82.511.829.668.634.0
2 Maharashtra16,050333.14.12.92.428.755.977.347.6
1 Punjab15,800113.64.22.61.96.019.370.040.4
Fourteen major states
Weighted average10,4102.52.62.31.9
National average13,0482.93.32.62.128.651.365.38
Standard deviation5,4150.671.050.670.4411.5111.5710.3613.27
Coefficient of variation10.270.400.280.230.400.230.16
Sources: Economic and Political Weekly States database; and IMF staff calculations.
Sources: Economic and Political Weekly States database; and IMF staff calculations.

Fact 1: The Gap Between Income Levels Across States Is Widening

The gap in per capita income between the richer and poorer states has widened over the previous three decades.1 Rich states have also grown over three times faster than poorer states so that by March 2004, the ratio of per capita income in the richest state (Punjab) to that in the poorest state (Bihar) had risen to 4.5 from 3.4 in 1970. There is also a strong correlation between the pace of growth and initial income levels. Dividing the sample into states that grew at rates above and below the national growth rate, we find that all of the poor states plus Kerala grew more slowly than the national average over the 1970–2004 period (Table 2.2). The rapidly growing states, which included mainly the middle-income states (Andhra Pradesh, West Bengal, and Karnataka) and the rich states (Gujarat and Maharashtra), grew over twice as fast as the slowly growing states.

Table 2.2.Absolute Divergence in Growth Rates
States Classified byAverage Growth in Real Per Capita Income
Real Per Capita Income in 19701970–20041970–791980–891990–992000–04
Poorest states0.46−3.41.92.91.8
Middle-income states2.740.82.25.65.0
Richest states2.683.04.33.34.5
1970–2004
Fastest growing states3.071.93.85.57.7
Slowest growing states1.430.62.62.61.8
National average2.610.23.43.54.4
Sources: IMF staff calculations; and Economic and Political Weekly States database.
Sources: IMF staff calculations; and Economic and Political Weekly States database.

Fact 2: Richer and Faster Growing States Are Generally Better in Reducing Poverty

A state’s record in reducing poverty reflects both differences in the level of growth and in the effectiveness of this growth in reducing poverty.

Overall, growth in India has reduced poverty less than proportionately and there is huge variation across states in poverty-growth elasticities (Table 2.3). Notably, not only have richer states tended to grow more rapidly, but they have been about 50 percent more effective in reducing poverty than poorer states for each percentage point of growth.2

Table 2.3.Cross-State Variation in Poverty-Growth Elasticities, 1977–20011
Contribution to Reduction in Poverty Rate of
Poverty-Growth Elasticity (B) 1977–2001Standard ErrorsState poverty-growth elasticity relative to India averageState growth rate relative to India average
Poorest states−0.68
Bihar−0.430.05***−0.41−0.18
Orissa−1.060.11***0.45−0.87
Uttar Pradesh−0.680.08***−0.07−0.36
Madhya Pradesh−0.390.05***−0.46−0.02
Rajasthan−0.820.05***0.12−0.31
Middle-income states−0.98
West Bengal−0.960.08***0.31−0.07
Andhra Pradesh−0.800.06***0.090.23
Kerala−1.540.09***1.09−0.04
Karnataka−0.630.04***−0.150.15
Richest states−1.02
Tamil Nadu−1.020.04***0.390.16
Haryana−0.730.15***0.00−0.03
Gujarat−1.010.07***0.380.05
Punjab−1.680.13***1.30−0.32
Maharashtra−0.620.05***−0.150.01
National average−0.730.03***
Coefficient of variation−0.51
Source: IMF staff calculations.

Head count poverty ratio regressed on real per capita net state domestic product. White-corrected heteroskedasticity errors. All variables are in logs. *** implies significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level.

Source: IMF staff calculations.

Head count poverty ratio regressed on real per capita net state domestic product. White-corrected heteroskedasticity errors. All variables are in logs. *** implies significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level.

To evaluate more closely the relative importance of differences in growth rates and poverty-growth elasticities, the decline in the poverty head count ratio is expressed as:

where pst is the reduction in the poverty rate between 1977 and 2001 in a given state, s, ɛ is the average India poverty-growth elasticity, g is the average India growth rate, and ɛs and gs are the state-specific average poverty-growth elasticity and growth rates. The first term measures the average reduction in poverty, the second differences across states in the effectiveness of growth in reducing poverty, and the third differences in growth rates across states. Terms two and three can be used to classify states according to the relative importance of the pace of growth and the effectiveness of this growth in reducing poverty.

States in the upper left-hand corner of Table 2.4 have the best of both worlds: growth in these states was faster than the Indian average and was effective in reducing poverty. In contrast, states in the lower right-hand corner experienced below-average growth and its effectiveness in reducing poverty was less than average. However, Table 2.4 underscores that some fast growing states were not as effective as slower growing states in reducing poverty. The policies explored in the next section may also help explain why some states are more effective than others in reducing poverty.

Table 2.4.Ranking of States by the Sources of Poverty Reduction, 1977–2001
High Growth1Low Growth1
High poverty elasticityAndhra PradeshKerala
GujaratOrissa
Tamil NaduPunjab
Rajasthan
West Bengal
Low poverty elasticityKarnatakaBihar
MaharashtraHaryana
Madhya Pradesh
Uttar Pradesh
Source: IMF staff estimates.

Using gross state domestic product. Most recent poverty data available are from 2001.

Source: IMF staff estimates.

Using gross state domestic product. Most recent poverty data available are from 2001.

Fact 3: Poor and Slower Growing States Generated Fewer Private Sector Jobs

While employment has risen across all states in the past three decades, the pace of job creation in middle- and high-income states far outstripped that of poorer states. India’s poorest and most populous states where about 40 percent of the population live only account for one-quarter of organized sector employment in India, although the poor quality of employment data should be kept in mind.3 While employment growth has in all states been driven by the public sector, the latter played a more crucial role in the poorer states where the private sector progressively shed jobs (Table 2.5).

Table 2.5.Structure of Employment in the Organized Sector1
1970/712001/02Annual Percentage ChangeEmployment Share 1970/71Employment Share 2001/02
PublicPrivateTotalPublicPrivateTotalPublicPrivateTotalPublicPrivatePublicPrivate
(In millions)(In percent of total employment)
Poorest states23.71.45.15.71.26.91.4−0.41.074.925.183.316.7
14 Bihar0.80.41.21.40.31.61.9−1.80.963.136.984.315.7
13 Uttar Pradesh1.40.51.91.70.52.20.7−0.50.472.028.079.021.0
12 Orissa0.30.10.40.70.10.82.50.42.281.019.089.210.8
11 Madhya Pradesh0.80.21.01.00.21.10.8−0.80.577.222.884.815.2
10 Rajasthan0.50.10.61.00.21.22.32.72.481.318.779.320.7
Middle-income states22.72.24.94.72.77.47.75.06.457.242.862.537.5
9 Andhra Pradesh0.70.31.01.50.62.12.32.32.370.829.271.328.7
8 West Bengal1.11.22.31.50.72.31.1−1.60.046.953.167.033.0
7 Kerala0.30.40.70.60.61.22.31.21.844.255.852.847.2
6 Karnataka30.60.30.91.10.81.91.93.12.466.833.258.741.3
Richest states23.42.55.95.73.69.29.88.59.260.040.062.038.0
5 Tamil Nadu0.90.71.51.60.92.51.91.11.657.742.364.135.9
4 Gujarat0.50.51.00.80.71.61.61.51.552.347.753.646.4
3 Haryana0.20.10.30.40.30.72.82.92.862.337.761.638.4
2 Maharashtra1.51.22.62.21.43.61.30.61.056.143.961.438.6
1 Punjab0.30.10.40.60.30.82.12.52.271.928.169.530.5
National average10.76.717.518.88.427.21.80.71.461.138.469.031.0
Share of poorest states34.620.729.130.314.425.4
Share of middle-income states25.232.728.025.131.827.2
Share of richest states31.537.233.530.342.134.0
Source: Ministry of Labour.

These states accounted for 90 percent of organized sector employment in 1970/71 and 86.5 percent in 2001/02.

Simple average over each income group.

1972/73.

Source: Ministry of Labour.

These states accounted for 90 percent of organized sector employment in 1970/71 and 86.5 percent in 2001/02.

Simple average over each income group.

1972/73.

States differ greatly in their ability to translate growth into jobs. Although puzzling for a labor-rich country, it has been well documented that growth in India has not been very job intensive—the national growth-employment elasticity for the organized sector is only 0.5. Moreover, poorer states have fared even worse than this. Notwithstanding this, Tables 2.6 and 2.7 show that high-growth states have generally been more successful in translating growth into jobs, which may also help explain why states such as Andhra Pradesh, Gujarat, and Tamil Nadu have made large inroads into poverty. However, it is also the case that some rapidly growing states (Madhya Pradesh, Maharashtra, and West Bengal) have been less successful in generating job-intensive growth.

Table 2.6.Cross-State Variation in Employment-Growth Elasticities, 1970/71–2001/021
Employment Growth ElasticityStandard Errors
TotalPublicPrivateTotalPublicPrivate
Poorest states20.650.770.13
Bihar30.320.52−0.460.06***0.10***0.08***
Orissa1.411.620.180.20***0.23***0.13
Uttar Pradesh30.410.59−0.160.08***0.10***0.06***
Madhya Pradesh30.280.310.100.09***0.10***0.05*
Rajasthan0.830.800.970.12***0.11***0.12***
Middle-income states20.440.510.37
West Bengal−0.100.22−0.550.02***0.07***0.06***
Andhra Pradesh0.680.690.650.05***0.07***0.04***
Kerala0.450.590.310.09***0.13***0.06***
Karnataka0.730.551.050.04***0.06***0.05***
Richest states20.580.610.55
Tamil Nadu0.490.550.390.05***0.08***0.03***
Haryana0.780.840.690.07***0.08***0.07***
Gujarat0.480.500.450.05***0.07***0.04***
Punjab0.810.750.980.04***0.05***0.03***
Maharashtra0.330.400.230.04***0.05***0.02***
National average0.500.590.310.052***0.075***0.018***
Coefficient of variation0.700.731.50
Source: IMF staff calculations.

Using employment in the organized sector. *** implies significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level.

Simple average across states in each income group.

Regressions included a dummy variable to capture year when new state was formed.

Source: IMF staff calculations.

Using employment in the organized sector. *** implies significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level.

Simple average across states in each income group.

Regressions included a dummy variable to capture year when new state was formed.

Table 2.7.Ranking of States by Sources of Employment Generation, 1970/71–2001/02
High GrowthLow Growth
Total employment
High employment elasticityAndhra PradeshOrissa
HaryanaRajasthan
Karnataka
Punjab
Low employment elasticityWest BengalBihar
GujaratKerala
Madhya PradeshUttar Pradesh
Maharashtra
Tamil Nadu
Private employment
High employment elasticityAndhra PradeshRajasthan
Gujarat
Haryana
Karnataka
Punjab
Tamil Nadu
Low employment elasticityMadhya PradeshBihar
MaharashtraKerala
West BengalOrissa
Uttar Pradesh
Source: IMF staff estimates.
Source: IMF staff estimates.

Fact 4: Capital and Labor Flows Do Little to Address Imbalances Across States

Economic activity is highly concentrated, and India’s most populous states contribute less to output than their share in the population. The five poorest states with 40 percent of the population produce only one-quarter of total output. The richest five states, which are home to only about one-fourth of India’s population, produce over 40 percent. There are also large geographical disparities in the sectoral distribution of economic activity. While about half of total agricultural value added in India is produced in the northern and central states, the coastal states of Maharashtra, Gujarat, and Tamil Nadu produced 40 percent of industrial and service sector output. Such disparities are not unusual. In fact, the concentration of economic activity observed in India is very similar to that observed by Easterly and Levine (2002) in the United States. The correlation between poverty and geographic location is also high. India’s poorest states are mainly located in the central and northern regions, where the head count poverty ratios generally exceed 30 percent. Middle- and high-income states are located mainly in the coastal areas.4

One might expect that capital—and jobs—would tend to move to poorer states, attracted by a pool of low-paid or unemployed workers. However, the evidence shows that capital goes primarily to the richer states, exacerbating the plight of poor states. Using the stock of credit from scheduled commercial banks to proxy capital stock, we find that the five richest states received a disproportionate share of capital, about 55 percent of total stock. The five poor states received only 15 percent.5 Chapter 5 also shows that about half of total foreign direct investment (FDI) approvals in India go to five rich states.

Another possible mechanism for equilibrating incomes across states would be labor migration from poorer to richer states. In fact, labor in India does migrate to the richer states, but the overall level of labor mobility in India across state borders is very low and does little to assist the convergence process (Table 2.8). Only 6 percent of migration in rural areas and 20 percent of migration in urban areas occurred across state borders.6 Net outward migration is highest from the northern and central states of Bihar, Uttar Pradesh, and Punjab. Delhi and the coastal states of Maharashtra and Gujarat are the prime migration destinations. India’s wealthiest states attracted about half of the total number of migrants during 1999–2000. However, limited cross-state migration is consistent with Cashin and Sahay (1996), who find that state-to-state migration in India is not very responsive to cross-state income differentials. The low level of cross-state migration may reflect language barriers and poverty, as poorer individuals may find it difficult to finance a move to a different state in the absence of family ties.

Table 2.8.Interstate Migration, 1971–2000

(In percent)1

Net Annual Migration Rate
1971198119912000
Poorest states
Bihar−0.00112−0.00105−0.00030−0.01862
Orissa0.000360.000280.000600.00233
Uttar Pradesh−0.00114−0.001750.00250−0.00458
Madhya Pradesh0.000940.000280.003600.00651
Rajasthan−0.00095−0.000470.00350−0.00351
Average−0.00038−0.000540.00198−0.00358
Middle-income states
West Bengal0.000860.000570.000900.00623
Andhra Pradesh−0.00035−0.000290.004300.00035
Kerala−0.00177−0.00128−0.00320−0.00104
Karnataka0.000350.000130.00090−0.00434
Average−0.00023−0.000220.000730.00030
Richest states
Tamil Nadu0.00103−0.00060−0.00010−0.00308
Haryana0.000870.00069−0.000400.03662
Gujarat0.000340.00053−0.001100.00857
Punjab−0.00207−0.00018−0.003200.00827
Maharashtra0.001810.002260.005700.02032
Delhi0.021660.02293
Average0.003940.004270.000180.01414
Sources: National Sample Survey Organisation (2001); and Cashin and Sahay (1996).

Average annual net migration as a share of state population at the start of each decade.

Sources: National Sample Survey Organisation (2001); and Cashin and Sahay (1996).

Average annual net migration as a share of state population at the start of each decade.

Table 2.9.Volatility in Economic Growth
States Classified by Real Per Capita Income in 1970Coefficient of Variation in Real Per Capita Income Grow
1970–20041970–791980–891990–992000–01
Poorest states4.119.142.422.012.78
Middle-income states2.927.581.801.230.74
Richest states2.745.742.481.080.99
1970–2004
Fastest growing states3.387.462.751.501.23
Slowest growing states1.894.911.020.840.96
National average1.383.550.580.600.55
Sources: IMF staff calculations; and Economic and Political Weekly States database.
Sources: IMF staff calculations; and Economic and Political Weekly States database.

Fact 5: Growth Has Been the Most Volatile in the Poorest States

Individuals, in particular the poor, are vulnerable to large swings in income. In this context, a key fact is that growth has been the most volatile in the poorer states, and increasingly so since the early 1980s. This stands in marked contrast to the experiences of rich and middle-income states. However, Table 2.9 shows that the fastest growing states (the three middle-income and two high-income states) experienced greater volatility in growth rates than slower growing states, suggesting that despite experiencing temporary busts, on average, these states ended up with higher per capita incomes.7

In sum, the stylized facts in this section suggest that the income gap between richer and poorer states has widened. States differ greatly in their ability to attract investment and translate growth into more jobs and less poverty. In many ways, these findings contrast with those of the theoretical neoclassical convergence literature, which predicts that states that are initially poor should grow faster than richer ones, and that capital and labor will migrate to ensure convergence.8 However, the concentration of economic activity across states may reflect other factors highlighted in the economic geography literature, such as locational advantages in terms of access to markets and supply sources (Redding and Venables, 2004), transport and congestion costs (Krugman, 1991), and scale economies and spillovers of knowledge and information (Fujita, Krugman, and Venables, 1999) that can lead to the agglomeration of economic activity. Moreover, the analysis also suggests that while high growth is generally associated with poverty reduction and job creation, growth alone is not enough to ensure good outcomes on these two fronts. We turn next to examine whether differences in economic policies across states affect economic performance.

Do Policies Matter?

There have been various attempts to assess econometrically whether differences in economic policies across states account for the differences in the pattern of state-level growth. Generally, states that liberalize factor markets and promote good institutions are found to have fared better than others. Besley and Burgess (2000, 2004) look at the impact of specific economic reforms on manufacturing and agricultural growth. They find that states that amended labor laws in favor of workers experienced lower growth in output, employment, investment, and productivity in the formal manufacturing sector and increases in urban poverty. In agriculture, states that amended land laws to encourage redistribution of land to laborers and the amalgamation of farms into viable units experienced higher investment, productivity, and output growth. Banerjee and Iyer (2005) use district-level data and find that areas in which proprietary land rights were historically given to landlords had significantly lower agricultural investments and productivity after independence than areas in which these rights were given to cultivators. Burgess and Pande (2004) noted that the Indian rural bank branch expansion program of 1977–90 significantly lowered rural poverty and increased nonagricultural output.9Kochhar and others (2006) found that states with weaker institutions and poorer infrastructure experienced lower GDP and industrial growth, particularly in electricity- and infrastructure-intensive sectors. Surveys of over 2,000 business establishments across 20 states conducted by Indicus Analytics (2004) are also suggestive of a positive relationship between a state’s economic policy environment and its growth performance.

To test this link more formally, we identify various time-series indicators of economic policy at the state level on the basis of the literature and the availability of data. The purpose is not to identify an exhaustive list of the determinants of growth or to rank the importance of each factor but rather to assess whether policies are linked with growth and whether they can account for the cross-state pattern of economic performance. In general, the real per capita growth rate of a state is related to two kinds of variables. The first type proxies initial economic conditions, such as the structure of a state’s economy. In line with the convergence literature (Barro and Sala-i-Martin, 1999), the second group of policy variables reflects actions by the government or individuals that can have a direct effect on a state’s steady-state or long-run level of per capita income. The extent of time-series data varies, but generally cover the 1973–2003 period with the exception of infrastructure where data are only available from the 1980s onward. The variables used in the analysis are described below and key correlations are illustrated in Figure 2.1.

Figure 2.1.State-Level Economic Growth and Changing Business Climate Indicators

Sources: Besley and Burgess (2000 and 2004); Reserve Bank of India; Ministry of Power; Department of Road Transport and Highways; National Sample Survey Organisation; and IMF staff calculations.

Initial Conditions

  • Initial income. If there is convergence, states with higher levels of income will tend to grow at a slower rate. The initial level of per capita income is measured using real NSDP and the coefficient on this variable is used to derive the rate of convergence.
  • Economic structure. States whose economic structure is more biased toward agriculture are expected to grow more slowly reflecting the low productivity of the largely subsistence sector. The economic structure of a state is measured using the lagged ratio of agriculture and industry in a state’s NSDP.

Policy Variables

  • Private investment/financial intermediation. Absent national account data on capital stocks and investment by state, the real stock of private sector credit per capita is used as a proxy for capital investment. This variable also reflects the depth of financial intermediation. Levine, Loayza, and Beck (2000) find robust evidence that financial development in general can foster faster long-run growth by ameliorating information and transaction costs. Thus states with greater levels of investment and/or more developed financial systems should experience a more rapid pace of growth.
  • Level and quality of human capital. Following Barro and Sala-i-Martin (1999), the stock of human capital is proxied using female literacy rates.10 Using female rather than overall literacy can also serve as a crude proxy for the quality of education in a state, with states that place greater emphasis on female literacy being viewed as more progressive. Moreover, there are also numerous channels through which female education can have an impact on the rate of economic growth. Empirical studies show that the education of mothers improves the education, nutrition, and health of their children. Education of women can improve the education prospects and standards of the next generation. Thus, this variable serves both as a control for differences in initial levels of human capital stock across states—convergence suggests that states with initially high levels of education would tend to experience lower growth rates—and the dynamic impact of education on growth.
  • Size of government. Cross-country studies of the determinants of economic growth generally find that countries with smaller governments had better growth performance (e.g., Easterly and Levine, 2002). Here the size of government is measured using the ratio of total state government expenditure to NSDP.
  • Industrial relations climate. States facing fewer labor disputes are likely to attract greater investment and this can spur growth. The analysis uses the lagged number of person-days lost to strikes and lockouts scaled by total organized sector employment to capture such effects.
  • Reform of labor regulations. State-level legislation that offers greater protection of workers and curtails the flexibility of employers to hire, fire, and organize their work practices may reduce productivity and deter investment. Besley and Burgess (2004) construct a measure that summarizes how industrial relations regulation in Indian states changed between 1947 and 1992, and this measure is extended here to include amendments implemented post-1992 reported in Malik (2003). State-level amendments to the 1947 Industrial Disputes Act are coded so that pro-worker amendments receive a score of one, pro-employer amendments score minus one, and changes that are neutral score zero. The scores are then accumulated over time to give a continuous quantitative picture of how the labor relations environment evolved. The method classifies Andhra Pradesh, Karnataka, Kerala, Madhya Pradesh, Rajasthan, and Tamil Nadu as pro-employer states. Gujarat, Maharashtra, Orissa, and West Bengal are “pro-worker” states. India’s six other large states did not implement any amendments to the Industrial Disputes Act over the period.
  • Infrastructure. States with more extensive transport networks should be better able to facilitate economic activity and attract investors. The penetration of transport networks is measured as the number of kilometers of roads scaled by the area of the state. Likewise, states with better power networks, as measured by transmission and distributional losses of state electricity boards, should be more attractive investment locations. This variable has also been interpreted as a proxy of state reform credentials by Kochhar and others (2006), where improvements in this ratio reflect the willingness of state governments to control losses of power from their network because of theft and unwillingness to charge users.

In contrast to other studies, the use of time-series data on these variables allows the analysis to assess whether changes in policies in a given state, as well as differences in the policy environment across states, affected cross-state growth rates. Our richer database may account for differences in the findings with other studies such as Ahluwalia (2002) and Kochhar and others (2006), who rely on time-invariant or static measures of state institutions sampled at fixed points in time.

The econometric analysis utilizes a generalized method of moments (GMM) dynamic panel estimate to assess the relationship between policy and economic growth. The panel consists of data for India’s 15 largest states for 1973/74–2002/03 averaged over six nonoverlapping five-year periods. The GMM estimator has the advantage that it allows past realizations of the dependent variable to affect its current level using lagged levels of the dependent and predetermined variables.11 Time dummies are included to account for time-specific effects. Robust standard errors are reported.

The results of the analysis are summarized in Table 2.10. Specifications I and II examine the question of absolute convergence, namely whether poor states grow faster toward their steady state than richer states absent policy controls. Specification II interacts initial income with a post-1991 dummy variable to assess if convergence or divergence accelerated post-1991. Specifications III-V examine the relationship between growth and state-level policies. Specification III includes all those variables for which data are available from 1973/74 to 2002/03, specification IV adds indicators of infrastructure that are available over a shorter time span, while specification V assesses whether state-level policies mattered more for individual economic performance post-1991 by interacting the policy variables with the post-1991 dummy. In sum, the findings suggest the following:

Table 2.10.India: Determinants of State Real Per Capita Income
Dependent Variable: Five-Year Average

Real Per Capita NSDP, 1973/M–2002/031
Absolute convergenceState policies and economic structure
Full sampleFull samplePost-1990s
IIIIIIIVV
Constant0.07**0.030.08***0.06***1.18*
(0.029)(−0.0424)(0.0131)(0.0213)(0.6965)
Initial conditions
In initial real per capita income0.59***0.58***0.42***0.38***0.08
(0.218)(0.208)(0.160)(0.082)(0.088)
In initial real per capita income* post-1990s0.02**
(0.009)
In agriculture share of net state domestic product (NSDP) lagged−0.18−0.27***−0.39*
(0.069)(0.069)0.216
In industry share of NSDP lagged−0.12*−0.02**−0.36***
(−0.070)(0.082)(0.102)
Policy variables In investment
0.12**0.090.12**
(0.053)(0.072)(0.062)
In female literacy rate−0.090.03−0.09
(0.061)(0.059)(0.104)
In size of government−0.32***−0.25***−0.16
(0.072)(0.062)(0.140)
In person-days lost to dispute per worker, lagged−0.02*−0.03***0.06***
(0.012)(0.012)(0.021)
Index of labor regulation, lagged0.01***0.01***−0.01***
(0.002)(0.003)(0.005)
Controls for state infrastructure Roads per km2
0.00
(0.058)
Transmission and distributional losses of electricity in percent of availability−0.17***
(0.064)
Convergence coefficient0.01***0.01***0.02***0.01***0.16*
Half-life49.0112.544.861.64.4
Time controlsYesYesYesYesYes
Serial correlation test (p value)0.580.470.570.750.29
No. of observations6060606060
Source: IMF staff estimates.

Robust standard errors are reported in parentheses. All variables, with the exception of the variable for labor regulation are in logs. *** implies significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level.

Source: IMF staff estimates.

Robust standard errors are reported in parentheses. All variables, with the exception of the variable for labor regulation are in logs. *** implies significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level.

  • Absolute convergence occurs very slowly. The coefficient on lagged income in specifications I and II is significant and suggest that initially poor states grow faster than initially rich ones—so-called absolute convergence—absent controls for differences in policies and economic structure. However, the rate of convergence is only about 1½ percent a year, which implies that it takes almost 50 years to close half the gap (also known as the half-life) between any state’s initial level of per capita income and its steady-state level of income. The coefficient on the interaction term for the post-1991 period suggests incomes continued to converge post-1991 but at an even slower pace than for the full period.
  • The differences observed in state incomes reflect wide gaps in their steady-state or long-run level of income. Specifications III-V find evidence of conditional convergence. In other words, poor states grow faster than rich states once controls that proxy for differences in the policies and economic structure are held constant. But the convergence coefficient in specification III changes only marginally relative to specifications for absolute convergence and the pace of convergence is broadly in line with the findings from other international studies (see Box 2.1).12 This suggests that differences in an individual state’s steady-state or long-run income potentials—rather than specific policies—have been the main drivers of the disparities in growth performance observed in India since 1970. However, in specification V, the speed of conditional convergence increases sharply and the half-life is reduced to about 4½ years. This suggests that differences in policies implemented by states in the 1990s have became important determinants of a state’s growth. This may reflect the fact that, with the move toward greater decentralization, a wider variety of policy approaches have been tried. Moreover, with greater openness to the global economy, good policies may be having a large potential payoff.
  • State-level policies have long-run growth effects. Greater investment—as measured by the stock of real private credit per capita—leads to economic growth.13 The quality of a state’s infrastructure also appears to be an important determinant of growth. While specification II does not find any significant relationship between growth and road penetration, rising transmission and distributional losses in the electricity sector adversely affect a state’s growth performance. On the other hand, the size of government adversely affects state-level growth suggesting that a shift in state spending from consumption to infrastructure investment could have a growth dividend. Specification V, which interacts the key policy variables with a post-1991 dummy, yields broadly similar results, although the increase in the magnitude of the coefficient on many of the policy variables suggests the policy environment of individual states became more important after 1991.
  • The impact of initial economic conditions can linger for long periods. Specifications II–V find that states with a greater initial dependence on agriculture or industry grew more slowly.
  • It is difficult to disentangle the impact of labor market policies on growth. Specifications III and IV suggest that, as expected, the number of person-days lost to labor disputes in the preceding period had an adverse impact on growth. However, the result that states that enacted pro-worker legislation experienced a better growth performance is puzzling. This may be a product of the fact that legislative changes may be a poor proxy for actual labor market flexibility because some rapidly growing states have chosen to enforce loosely or even exempt firms from such provisions. In fact, the results on labor market conditions are driven by one outlier, West Bengal, a state which has been far more active than others in enacting pro-worker amendments to the Industrial Disputes Act but which has exempted many key sectors from such provisions. Once West Bengal is excluded, the coefficients on the two labor market variables become insignificant in all specifications.
  • Female literacy is not found to have a significant exogenous impact on states’ growth performance. In fact, the coefficient suggests a negative relationship with growth, a result that is shared with many other such studies in this field (see, for example, Barro and Sala-i-Martin, 1999, and Kalaitzidakis and others, 2001). Barro and Sala-i-Martin argue that female education is picking up standard conditional convergence effects whereby states with lower initial human capital grow faster given their greater distance to their steady state. Szulga (2005), on the other hand, argues that the estimate of the impact of female education on growth is biased because many educated females do not enter the labor force. Even when aggregate literacy levels are used in place of female education, human capital is again found to have a negative, albeit insignificant, effect on growth, a finding that is also confirmed by Islam (1995) in cross-country growth regressions.

Box 2.1.International Evidence on Regional Convergence

The speed of conditional convergence in incomes across Indian states mirrors that found by studies of regions in other industrial and emerging market economies. The speed of convergence across regions in developing countries or in panel data sets generally appear to be faster than that commonly found in studies of regions of industrial countries or across countries. For example, Barro and Sala-i-Martin (1991) and Sala-i-Martin (1996) show that the speed of convergence across the regions of the United States, Europe, and Japan is close to 2 percent. Islam (1995) finds that speeds of convergence across 97 countries range from 4 percent to 10 percent, depending on the method of estimation used. Canova and Marcet (1995) find that the speed of convergence across regions of Western Europe to be as high as 20 percent, whereas Caselli, Esquivel, and Lefort (1996) estimate a speed of convergence of about 13 percent a year across 97 different countries.

China. Jian, Sachs, and Warner (1996) examine convergence in incomes across China’s provinces. They find only weak evidence of convergence during the central planning period (1952–65), and during the cultural revolution (1965–78) incomes across provinces diverged strongly. Convergence following the start of economic reforms in 1978 was most closely associated with rural reform, and was strong in coastal areas where trade and investment flows were liberalized. However, their study only extended to 1993 and toward the end of their sample there was some tendency toward divergence. Using a GMM estimator, Weeks and Yao (2003) show that China’s provinces converged at a rate of only 0.4 percent a year in the 1953–77 period but in the postreform 1978–97 period convergence accelerated to 2.2 percent a year.

Korea. Koo, Kim, and Kim (1998) note that per capita incomes across Korea’s 10 states converged between 1967 and 1992 at an annual rate of between 4 percent and 6 percent. However, in two five-year subperiods between 1972 and 1982, income diverged because regions responded differently to the 1970s’ oil price shocks. However, industrial policy promoted convergence during 1977–82. Migration was found to have little impact on regional convergence.

Latin America. In Brazil, Ferreira (1999) finds evidence of conditional convergence between 1939 and 1995 and estimates that by 1995 the income of a number of poor states was very close to their steady-state values, suggesting that looking forward large income disparities would remain across states. In Colombia, Cárdenas and Pontón (1995) determine a rate of convergence across Colombia’s 22 departments between 1950 and 1990 of 4 percent a year without controls for initial conditions, and 5¼ percent a year if regional controls were included in the analysis. Labor migration did not play a large role in promoting convergence, except in the 1960s. Elías and Fuentes (1998) find evidence that rates of conditional convergence across regions were higher within Chile than in Argentina between 1960 and 1985. After controlling for differences in initial conditions, they estimate the rate of convergence in labor income to be 2 percent.

Spain. De la Fuente (2002) finds evidence that the speed of convergence across Spanish regions varied over time and was not necessarily fastest during periods of high national growth. Income per capita converged at an average annual rate of about 2½ percent between 1965 and 1975, slowed to about 1 percent during the crises of 1975–85, and fell to 0.4 percent between 1985 and 1995, at a time when Spain was growing faster than most industrial countries. The slowdown in the rate of convergence reflects lower employment generation and a fall in internal migration. Leonida and Montolio (2004) note that the convergence in incomes stalled in the 1980s but recommenced in the 1990s.

In sum, the findings suggest that states can influence their relative growth performance by adopting better economic policies. A state can improve its long-run economic position by bringing about improvements in its investment, fiscal, and infrastructure policies. The results on the impact of state economic structure also point to a need in some poor performing states to diversify economic activity away from agriculture and industry and adopt policies that make these sectors more productive.

Conclusions

This chapter examined how growth and economic performance have varied across India’s largest states over the past 30 years. It documented five stylized facts about their performance: (1) the gap in real per capita incomes between rich and poor states has widened over time; (2) rich and faster growing states have generally been more effective in reducing poverty; (3) poor and slower growing states have had very little success in generating private sector jobs; (4) labor and capital flows appear to do little to close the gap in incomes between poor and rich states; and (5) poor states experience the greatest volatility in economic growth.

This chapter then examined the link between state-level policies and economic growth. The econometric analysis presented evidence that state-level polices are a key factor influencing the pattern of economic growth across Indian states. Greater private sector investment, smaller governments, and better state-level institutions (as proxied by transmission and distributional losses of state electricity boards) are found to be positively associated with growth performance, but the impact of labor market policies are more difficult to discern. The historical structure of economic activity in a state also appears to matter for a state’s subsequent growth performance. All this suggests that states can influence their relative growth performance and accelerate convergence through their policy choices.

Appendix. Data Description

State-level income data is derived from the Economic and Political Weekly States database and the Central Statistical Organisation. The sample of 15 states account for 95 percent of India’s population and about 80 percent of its domestic product. Using these data we construct an annual series on real net state per capita incomes and the share of agriculture, industry, and services by splicing the three base year series on real NSDP to arrive at a series based in 1993/94 prices. In the absence of state-level aggregate investment or capital stock data, we utilize the stock of credit extended by scheduled commercial banks reported in the Reserve Bank of India’s (RBI) Basic Statistical Tables from the banking system starting in 1973, translated into real terms using state-level NSDP deflators. Literacy rates are derived from various rounds of the National Sample Survey with intervening survey years constructed by linear extrapolation. Data on labor market regulation were provided by Tim Besley, and are available at http://sticerd.lse.ac.uk/eopp/research/indian.asp, and were updated using Industrial Law(Malik, 2003). Employment and labor disputes data were provided by the Ministry of Labour, as reported in the annual editions of the Indian Labour Yearbook. Electricity sector transmission and distribution losses as a percent of availability were derived from the Annual Reports of State Electricity Boards available from the Ministry of Power and Planning Commission (see http://planningcommission.nic.in/reports/genrep/reportsf.htm). Data on state government spending were derived from the World Bank’s States Fiscal Database and http://sticerd.lse.ac.uk/eopp/research/indian.asp, and the primary source for these data is the RBI’s annual report on state finances.

References

    AhluwaliaMontek S.2002“State-Level Performance Under Economic Reforms in India” in Economic Policy Reforms and the Indian Economyed. by Anne O.Krueger (New Delhi: Oxford University Press).

    • Search Google Scholar
    • Export Citation

    AiyarShekhar2001“Growth Theory and Convergence Across Indian States: A Panel Study” in India at the Crossroads: Sustaining Growth and Reducing Povertyed. by TimCallenPatriciaReynolds and ChristopherTowe (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    BajpaiNirupam and Jeffrey D.Sachs1996“Trends in Inter-State Inequalities of Income in India,”Harvard Institute for International Development Development Discussion Paper No. 528 (Cambridge, Massachusetts: Harvard University).

    • Search Google Scholar
    • Export Citation

    BanerjeeAbhijit and LakshmiIyer2005“History, Institutions, and Economic Performance: The Legacy of Colonial Land Tenure Systems in India,”American Economic ReviewVol. 95No. 4 (September) pp. 1190213.

    • Search Google Scholar
    • Export Citation

    BarroRobert and XavierSala-i-Martin1991“Convergence Across States and Regions” Brookings Papers on Economic Activity: 1 pp. 10758.

    • Search Google Scholar
    • Export Citation

    BarroRobert and XavierSala-i-Martin1999Economic Growth (Cambridge, Massachusetts: MIT Press).

    BesleyTimothy and RobinBurgess2000“Land Reform, Poverty Reduction and Growth: Evidence from India,”Quarterly Journal of EconomicsVol. 115No. 2 (May) pp. 389430.

    • Search Google Scholar
    • Export Citation

    BesleyTimothy and RobinBurgess2004“Can Labor Regulation Hinder Economic Performance? Evidence from India,”Quarterly Journal of EconomicsVol. 119No. 1 (February) pp. 91134.

    • Search Google Scholar
    • Export Citation

    BhattacharyaB.B. and S.Sakthivel2004“Regional Growth and Disparity in India: Comparison of Pre- and Post-Reform Decades,”Economic and Political WeeklyVol. 39No. 10 pp. 107177.

    • Search Google Scholar
    • Export Citation

    BurgessRobin and RohiniPande2004“Do Rural Banks Matter? Evidence from the Indian Social Banking Experiment,”CEPR Discussion Paper No. 4211 (London: Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation

    CanovaFabio and AlbertMarcet1995“The Poor Stay Poor: Non-Convergence Across Countries and Regions,”Universitat Pompeu Fabra Economics Working Paper No. 137 (Barcelona).

    • Search Google Scholar
    • Export Citation

    CárdenasMauricio and AdrianaPontón1995“Growth and Convergence in Colombia: 1950–1990,”Journal of Development EconomicsVol. 47 (June) pp. 537.

    • Search Google Scholar
    • Export Citation

    CaselliFrancescoGerardoEsquivel and FernandoLefort1996“Reopening the Convergence Debate: A New Look at Cross-Country Growth Empirics,”Journal of Economic GrowthVol. 1No. 3 pp. 36389.

    • Search Google Scholar
    • Export Citation

    CashinPaul and RatnaSahay1996“Internal Migration, Center-State Grants, and Economic Growth in the States of India,”IMF Staff PapersVol. 43 (March) pp. 12371.

    • Search Google Scholar
    • Export Citation

    De la FuenteAngel2002“Regional Convergence in Spain, 1965–95,”CEPR Discussion Papers No. 3137 (London: Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation

    EasterlyWilliam and RossLevine2002“It’s Not Factor Accumulation: Stylized Facts and Growth Models,”Central Bank of Chile Working Paper No. 164 (Santiago).

    • Search Google Scholar
    • Export Citation

    ElíasV.J. and R.Fuentes1998“Convergence in the Southern Cone,”Estudios de EconomíaVol. 25No. 2 (December) pp. 17989.

    • Search Google Scholar
    • Export Citation

    FerreiraAfonso Henriques Borges1999“Concentraçao Regional e Dispersão das Rendas Per Capita Estaduais: Um Comentário,”Estudias EconomicãsVol. 29No. 1 (January–March) pp. 4763.

    • Search Google Scholar
    • Export Citation

    FujitaMasahisaPaulKrugman and Anthony J.Venables1999The Spatial Economy: Cities Regions and International Trade (Cambridge, Massachusetts: MIT Press).

    • Search Google Scholar
    • Export Citation

    Indicus Analytics2004Measuring Inter-State Differences in Investment Climate: A Report for the Twelfth Finance Commission (New Delhi).

    • Search Google Scholar
    • Export Citation

    IslamNazrul1995“Growth Empirics: A Panel Data Approach,”Quarterly Journal of EconomicsVol. 110No. 4 (November) pp. 112770.

    • Search Google Scholar
    • Export Citation

    JianTianlunJefferySachs and AndrewWarner1996“Trends in Regional Inequality in China,”NBER Working Paper No. 5412 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    KalaitzidakisPantelisTheofanis P.MamuneasAndreasSavvides and ThanasisStengos2001“Measure of Human Capital and Nonlinearities in Economic Growth,”Journal of Economic GrowthVol. 6No. 3 pp. 22954.

    • Search Google Scholar
    • Export Citation

    KochharKalpanaUtsavKumarRaghuramRajanArvindSubramanian and IoannisTokatlidis2006“India’s Pattern of Development: What Happened, What Follows?”IMF Working Paper No. 06/22 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    KooJaewoonYoung-YongKim and SangphilKim1998“Regional Income Convergence: Evidence from a Rapidly Growing Economy,”Journal of Economic DevelopmentVol. 23No. 2 (December) pp. 191203.

    • Search Google Scholar
    • Export Citation

    KrugmanPaul1991“Increasing Returns and Economic Geography,”Journal of Political EconomyVol. 99No. 3 pp. 48399.

    KumarSanjay2004“Impact of Economic Reforms on the Indian Electorate,”Economic and Political WeeklyVol. 39No. 16 pp. 162130.

    • Search Google Scholar
    • Export Citation

    LeonidaLeone and DanielMontolio2004“On the Determinants of Convergence and Divergence Processes in Spain,”Investigaciones EconomicasVol. 28No. 1 pp. 89121.

    • Search Google Scholar
    • Export Citation

    LevineRossNormanLoayza and ThorstenBeck2000“Financial Intermediation and Growth: Causality and Causes,”Journal of Monetary EconomicsVol. 46No. 1 pp. 3177.

    • Search Google Scholar
    • Export Citation

    MalikP.L.2003Industrial Law (Lucknow, India: Eastern Book Company).

    National Sample Survey Organisation2001Migration in India 1999–2000 (New Delhi: Ministry of Statistics and Programme Implementation).

    • Search Google Scholar
    • Export Citation

    RaoM.G.R.T.Shand and K.P.Kalirajan1999“Convergence of Incomes across Indian States: A Divergent View,”Economic and Political WeeklyVol. 34No. 13 pp. 76978.

    • Search Google Scholar
    • Export Citation

    ReddingStephen and Anthony J.Venables2004“Economic Geography and International Inequality,”Journal of International EconomicsVol. 62 (January) pp. 5382.

    • Search Google Scholar
    • Export Citation

    Sala-i-MartinXavier1996“The Classical Approach to Convergence Analysis,”Economic JournalVol. 106 (July) pp. 101936.

    ShiojiEtsuro1997“Convergence in Panel Data: Evidence from the Skipping Estimation,”Yokohama National University Working Paper (Yokohama, Japan).

    • Search Google Scholar
    • Export Citation

    SinhaTapen and DipendraSinha2000“No, Virginia, States in India Are Not Converging,”International Indian Economic Association Working Paper No. 2000-09-01 (September).

    • Search Google Scholar
    • Export Citation

    SzulgaRadek2005“A Dynamic Model of Female Labor Force Participation and Human Capital Investment” (unpublished; University of California, Department of Economics).

    • Search Google Scholar
    • Export Citation

    VisariaLeela and PravinVisaria2003“Long-Term Population Projections for Major States, 1991–2101,”Economic and Political WeeklyVol. 38No. 45.

    • Search Google Scholar
    • Export Citation

    WeeksMelvyn and James YudongYao2003“Provincial Conditional Income Convergence in China, 1953–1997: A Panel Data Approach,”Econometric ReviewsVol. 22No. 1 pp. 5977.

    • Search Google Scholar
    • Export Citation
1Using net state domestic product (NSDP). Gross state domestic product (GSDP) was not available for all states for this time period.
2For each state, we run the following regression:
where LNPOVERTYst is the Deaton-corrected state poverty head count ratio from the National Sample Survey, t is time (within-sample years are extrapolated), and LNGDP is real per capita NSDP. β is the elasticity.
3The organized or official sector comprises enterprises registered under the 1951 Industries Act and covers all enterprises that employed 100 workers or more and do not use electricity, or firms that employ 50 or more workers and use electric power. organized employment accounts for about 10 percent of the labor force, so the analysis in this section should only be treated as indicative of broad trends in employment. However, formal sector employment may also be preferable, in a policy sense, to work in the unorganized sector, as it tends to offer higher wages.
4The correlation between the head count poverty rate and a dummy variable that is set equal to one if the state is located in the central and northern regions is 0.83. If the dummy variable is set to capture coastal states the correlation turns negative (-0.35).
5Using the location where credit was disbursed may overstate the degree of spatial concentration if borrowers utilize the funds borrowed in financial centers, such as Mumbai, in a different state.
6Urban-to-urban and rural-to-urban each account for one-fifth of interstate migration.
7Growth rates are averaged over five-year periods to help smooth cyclical fluctuations. The volatility in income growth between 1970 and 2004 was over three times the variation in cross-state incomes. The cross-sectional standard deviation averaged about 0.5 percentage points in the past three decades, but standard deviation over time averaged 1.6 percentage points.
8In a steady-state framework, per capita growth rates vary inversely with the distance a state is from its own steady-state level of growth. A poorer state with a relatively low capital-to-labor ratio should enjoy higher rates of return on capital and therefore high growth rates, as it converges to its steady state, assuming a constant returns to scale technology.
9Under this program, a commercial bank was granted permission by the Reserve Bank of India to open a branch in a location with one or more bank branches only if it opened four branches in locations with no bank branches.
10Using overall literacy rates in lieu of female literacy rates did not alter the results.
11Ordinary least square estimates are inconsistent in the presence of a lagged dependent variable and fixed effects.
12While the rate of convergence is close to that found by Cashin and Sahay (1996) for India in the 1960s-1980s, it is far lower than that reported by Aiyar (2001), who found convergence occurred at a rate of about 20 percent a year. The high convergence coefficient in the latter study is most likely the outcome of the fixed-effects estimator where inclusion of lagged dependent variables can result in upward bias (see Shioji, 1997, who demonstrates that fixed-effects estimates can be biased upward by between 7 percent and 15 percent; also see Islam, 1995).
13Since the GMM panel estimator controls for endogeneity, the findings suggest that the exogenous component of the relevant policy variable, take for example here investment, exerts a positive impact on economic growth.

    Other Resources Citing This Publication