1. Introduction

Economic development has historically been associated with structural changes in the national economies. Economic development is sometimes defined as a process combining economic growth with changing share of different sectors in the national product and labor force. The most common structural changes that have been observed historically have followed a sequence of shift from agriculture to industry and then industrial sector to services sector. A developing economy is characterized by a predominant share of agriculture, but as the economy develops further, the share of industry increases and that of agriculture declines, and subsequently after reaching a reasonably high level of development, the services sector increases in importance, becoming a major component of the economy. This pattern has not only been observed historically, but also holds across the countries with different levels of development. Structural shifts and changing sectoral shares are found to hold both for the national product and the work force.

It is established stylized fact that the share of agriculture in GDP falls as the level of development rises and it increases with the increase in the shares of industry and service sectors. However, the world is changing fast and continuously, and the patterns of development do not need to hold forever. It is also unlikely that all these relationships and stylized facts of the last forty years will suddenly break down and become irrelevant. Theories need to continually investigate the patterns of development using new and country specific data to establish whether the relationships continue to hold. With increasing openness of economies and trade playing significant role in them, changes in demand pattern can be met through trade and countries can have a product pattern, very different from the pattern of consumption demand, largely based on comparative advantage.

By revisiting past models of structural transformation, in particular the work of Chenery (1960) on the patterns of sectoral transformation, this paper aims to establish new development patterns based on a revised theoretical approach and available data. A major conceptual contribution of Chenery’s work was the identification of factors which affect a country’s structural change, specifically: (a) The normal effect of universal factors that relate to the levels of income; (b) the effect of other general factors such as market size or natural resources over which the government has little or no control; (c) the effects of the country’s individual history, its political and social objectives, and the particular policies the government has followed to achieve these (Chenery and Syrquin, 1975:5)

The income effect includes both supply and demand effect. The demand effect is usually associated with the factor that rising income levels lead to change in the composition of demand, of which the decline in the share of food (Engel’s law) is the most notable feature. The supply effect, on the other hand, entails two factors of general importance: (1) The overall increase in capital stock per worker, and (2) the increase in education and skills of all sorts. Moreover, as the production in which labour, capital, and skills can be combined vary from sector to sector; the change in factor supplies causes a systematic shift in comparative advantage as per capita income rises (Chenery, 1960: 624-625).

The literature on structural change points that country’ size has significant effects on the patterns of sectoral transformation, because economies of scale, resource endowments, and scale of domestic demand often vary with country size (Chenery and Syrquin, 1975; Chenery and Taylor, 1968; Syrquin, 1988). Kessing and Sherk (1971) show that population density plays an important role on patterns of trade and development. Densely populated areas appear to have a greater impact, in particular, on increased exports of manufactured goods relative to primary products. This relationship would suggest that only the most densely populated, small, developing countries can expect early successful export specialization in manufacturing sector.

However, we know little about the pattern of agricultural, industrial and service sectoral growth in India given the changing world environment, for example, with the current high world prices for basic agricultural and industrial commodities. This study contributes to the existing literature gap by investigating the relationship between agriculture, industry and service sectors share in GSDP at constant prices of 2004-05 using data of thirty two states and UT’s of Indian economy. The major contribution of the present paper is establishing the relationship between shares of agriculture, industry, manufacturing and service sectors in GSDP and the level of development by utilizing state level data available for thirty two states and UT’s of Indian economy. Previous studies were mainly country cross sectional or country specific time series data based and the state level perspectives were not given due consideration. India is very divergent country to its economy and hence it is essential to conclude on basis of state specific panel data for recently high growth period and to know that what has happened new in this short period that growth trends have been started to decline.

We expected a significant negative relationship between the shares of agriculture in GSDP with NSDP per capita and significant positive relationship between the shares of industry sector & service sector with NSDP per capita. Understanding inter-sectoral linkages could shed important insights on the transition process, and such information should assist policymakers identify the optimal policies to accelerate economic growth in states of India.

Section 2 deals with the empirical literature on sectoral interlink age in India. Section 3 is devoted to the discussion of methodology followed by selection of data used for the present paper. Section 4 presents the results. Finally, summary of the study has been provided in the last section with some concluding remarks.

2. Literature Review

A number of researchers have studied sectoral interlink ages in the Indian economy. In one of the earliest studies on the subject, Rangarajan (1982) found strong degree of association between the agricultural and industrial sectors. In particular, it has been observed that an addition of 1% growth in the agricultural sector stimulates the industrial sector output to the extent of 0.5%.

Bhattacharya and Mitra (1997) provided empirical evidence in support of positive linkage among the broad sectors. It established that many service activities are significantly associated with the agricultural and industrial sectors and this helps in overall employment generation.

Kalirajan (2004) studies the pattern of economic growth for 15 major states in India and the contribution of these states on India's GDP. His research shows that growth patterns are different among these states. Only seven states show a consistent increase in growth. Industry-oriented states are seen to grow more rapidly and absorb more labor. The increase in investment and growth in the secondary sector have significant relationship with the GDP growth rates of these states.

Diaz-Bautista (2003) applies the techniques of co-integration and Granger causality to examine the causal relationship between industrial growth and overall economic performance in the Mexican economy. The empirical results indicate that the industrial sector and overall economy are co-integrated and have a long run relationship. A well-established body of theoretical and empirical researches supports the conclusion that industries are engines of growth.

Libanio (2006) investigates the relation between manufacturing output growth and economic performance from a Kaldorian perspective by estimating Kaldor’s first and second growth laws for a sample of seven Latin American economies during the period 1985-2001. The relation between industrial growth and GDP growth can be explained by the effects of manufacturing on productivity levels in the whole economy. The results of this study confirm manufacturing as the engine of growth, and suggest the existence of significant increasing returns in the manufacturing sector in the largest Latin American economies. In addition, the study also concludes that productivity growth seems to respond positively to output growth in the manufacturing sector during the period of analysis. Overall, the results confirm the existence of increasing returns in the manufacturing sector, and the possibility of cumulative growth cycles in the region, based on the expansion of industrial activities.

Elhiraika (2008) investigates the role of structural dynamics and transformation, especially in the form of increased manufacturing share in aggregate output, in accelerating growth and reducing growth volatility in Africa. The study uses data from 36 African countries to examine the key determinants of manufacturing share in aggregate output and its relationship with real GDP growth and growth volatility. The analysis indicates that an increased manufacturing share in total output has the potential to raise GDP growth and reduce growth volatility. Therefore, the study demonstrates that economic transformation through increased share of manufacturing value-added in aggregate output has the potential to accelerate growth and reduce growth volatility.

Blecker (2009) found that four external variables (the US growth rate, net financial inflows, real oil prices, and the one-year lagged real exchange rate) explain most of the annual fluctuations in Mexico’s growth during the period 1979–2007.

To determine the relationship between the services industry and economic growth in China, Wang and Li (2010) use an empirical analysis involving unit-roots, co-integration and Granger causality tests based on time series data from 1990 to 2008. The results indicate that the Granger causality and long-term stable equilibrium relationship between the services, industry and economic growth. In other words, the development of the services industry plays an important role in economic growth in China.

In addition, Linden and Mahmood (2007) analyze the long run dynamic relationship between sector shares (agriculture, manufacturing and services) and economic growth for 15 Schengen countries in the period 1970 to 2004. Using panel co-integration techniques, paper evidenced that the relationships between services-share growth and the growth rate of real per capita GDP are bi-directional. Hence, their study confirms that feedback impact exists between the services sector and the growth rate of real per capita GDP.

In conclusion, based on the discussion of previous researches, this study finds that manufacturing and industry sectors have positive relationship with economic growth. In other words, the development of these economic sectors will raise economic growth.

3. Methodology

The specifications for agriculture, industry and service sectors share in GDP variable is adapted from the principal specification of Chenery and Syrquin (1975), Syrquin and Chenery (1989) and Branson et al (1998):

(1)

Where,

X is the dependent variable, taken as the share of agriculture and agriculture subsectors in GDP,

Y is the income level measured as GNP per capita,

N is the country’s population density, and

F is the net resource inflow, measured as imports minus exports of goods and nonfactor services as a share of total GDP and ε is the error term.

While Chenery initially felt it was crucial to have a sufficiently large sample size and for each demand component to be a function of income level, he later adopted single functions of income and population instead.

(2)

This makes it possible to view the effects of income level and country size using a linear logarithmic regression equation to estimate the value added level. This Equation became the foundation for subsequent structural change research and its modifications have been used in later studies. For example Chenery and Taylor (1968) included a quadratic term for income as the decline in elasticities with rising income levels became apparent. In later years, Chenery and Syrquin adopted a more general equation as shown below, allowing a non-linear effect for population and including dummy variables to identify period effects (Chenery and Syrquin, 1975; Syrquin and Chenery, 1989):

(3)

Where X is a dependent variable covering different aspects of structural change (usually expressed as a share in GDP), y is per capita GNP, N is population in millions, and T is a dummy variable for time periods taking a non-zero value for different periods. Chenery, Robinson and Syrquin (1986) concluded that the patterns are, to some degree, robust to time trend, therefore, cross-country estimations ought to reflect “true pattern”. In equation 4, the specification of fixed cross section (country specific) effect was applied, which is equivalent to including dummy variables for countries.

(4)

On account of conceptual and econometrical problems with equation (2), there are compelling reasons to refine the model in two ways in order to explain the patterns of economic growth. First, due to the simultaneous determination of supply and demand, output and income variables are endogenous determined within the model. In such a case, the least square estimator applied by Chenery reveals biased and inconsistent results. We are using the panel data method, which can separate timeas well as country-specific effects from the coefficients of the variables included in the equations, is more appropriate than the cross-section, single-period approach adopted by Chenery (1960). Secondly, It has been observed that the conventional regression method for examining the relationship between two time series variables that are non-stationary will often lead to spurious regression (Harris, 1995)¹. Most economic series are non-stationary and contain one or more unit roots. Therefore, in order to establish the true relationship between time series variables, one must check for the non stationarity or presence of unit roots in these variables (Granger and Newbold, 1994 and 1977). For that we can exercise the standard Dickey-Fuller (DF) and Augmented Dickey-Fuller unit root test. Accordingly, if the variable has no unit root, then they are stationary and if it has unit roots, then differencing the variable could make it stationary. Hence, we proceed with testing for the unit root.

Given this situation and the limited goal of the present paper, we limit the analysis to regressions including time trend and three dummy variables showing high income states, EAG states and high density states, using annual data from 2004-05 to 2013-14 for thirty two Indian states and UT’s. Using dummy variable for time periods taking a non-zero value for different periods, an appropriate equation may be written as follows in Chenery-Syrquin frameworks:

(5)

Chenery, Robinson and Syrquin (1986) concluded that the patterns are, to some degree, robust to time trend, therefore, cross-country estimations ought to reflect “true pattern”. Due to limited time period in the present paper, it was found an appropriate to use group of states of similar characteristics as dummy variable. Using three dummy variables showing high income states, EAG states and high density states in the place state specific dummy variable, equation 5 can be rewritten as:

(6)

Finally, the specification of fixed cross section (country specific) effect was applied, A Chow test will be used to select between Pooled OLS and Fixed Effects estimation techniques. If the Chow Test suggests using Fixed Effects, a Hausman test will be used to select between Fixed Effects and Random Effects estimation. If Hausman test suggests employing Random Effects estimation, then this is confirmed against Pooled OLS using a Breusch and Pagan Lagrangian multiplier test.

4. Data

The data used in the analysis are obtained from Central Statistical Organization (CSO).Detail of selected variables and their definition has been provided in Table 1. Data on shares of agriculture and agriculture allied sector, industry sector, service sector, NSDP per capita, population size were all obtained from the same source. The data used in this study range from 2004-05 to 2013-14, a period when India registered average growth rates of 6.4 percent per annum. This period is also a period when India experienced relative stability in macroeconomic variables and political situation. Descriptive statistics of the variables have been provided in Table 2.

Table 1. Detail of Explanatory Variables

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View next table

Table 2. Descriptive Statistics of the Panel Variables

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

The agricultural sector consists of three broad segments- agriculture and allied activities, forestry &logging and fishing. The industrial sector covers four main segments- Mining and Quarrying, Manufacturing, Electricity, Gas & Water Supply and Construction. Service sector includes trade, hotels & restaurants, transport, storage & communication, financing, insurance, real estate & business services and community, social & personal services. NSDP per capita is used in this study as a proxy of the level of development for the state. This is commonproxy for development level that has been consistently used by previous studies(Branson et al 1989; Syrquin and Chenery 1989).The population density variable controls for country size and effects of economies of scale. Studies, such as Syrquin and Chenery (1989) and Branson et al (1998), used population density to control for effects of economies of scale because from a theoretical point of view, it is a better proxy for economies of scale than population size. Population density was calculated by dividing total population in a given year by the land area which is measured in square kilometers.

5. Empirical Results

At first stage of analysis, we went through a preliminary analysis of data by calculating the unconditional correlation among study variables. The correlation results (see Table 3) show the significantly high positive correlation between shares of service sector and per capita NSDP, while the Share of agriculture and allied sector is negatively correlated with per capita NSDP. It implies that the economic growth is strongly associated with higher shares of service sector in GSDP.

Table 3. Correlation coefficients, using the observations 1:1 - 32:9

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Table 4. Estimation Results Using Pooled OLS, 2004-05-2013-14.

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Table 5. Estimation Results Using fixed effects, 2004-05-2013-14.

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Table 6. Im-Pesaran-Shin and Levin-Lin-Chu Test for Panel Unit Root Test, 2004-2013

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

In this paper an attempt has been made to establish relationships among per capita NSDP (Income effect), population density (size effect) and sector shares in GSDP (Structural Change). Table 4 shows income and size coefficients and the goodness of fit of the regressions, as indicated by the adjusted R squares and the results are based on the foundation equation derived by Chenery (1960) for subsequent structural change research. Both the dependent and explanatory variables are expressed in logarithmic terms to measure the elasticity of each coefficient, which explains as expected amount of change in logarithm value of dependent variable (Sector share) for one unit change in logarithm value of independent variables adjusting for the effect of other variables. There are some conceptual and econometric problems with equation (3), as these results are biased and inconsistent shown by Durbin - Watson value. Some other studies like Harguchi, N. and Rezonja, G. (2010) suggested using panel data method, which can sepretate time as well as country-specific effects from the coefficients of the variables included in the equations. Table 4 shows results of fixed effects, explanatory power and panel diagnostic tests. If we compare the results derived in Table 4 and Table 5, we find that explanatory powers of the models have been increased significantly but the problem of spurious regressions cannot be tackled properly, as shown by DW. Hence, in the next step it was decided to find stationary of time series variables. The unit root results clearly suggest that all the variables attain level of stationary at first difference (see Table 6).

It is now left to determine the structural change across states in India after 2004, the period when India achieved comparative high growth. The regression results for the agriculture and allied sector, agriculture sector, industry sector, manufacturing sector, service sector and time trend, which are based on equation 5 are reported as follows:

(5.1)

(5.2)

(5.3)

(5.4)

(5.5)

Results show that an increase in the share of industry and manufacturing sectors has significant positive impact on economic growth, while any increase in the share of agriculture will decline the present level of per capita NSDP on constant 2004-05 prices. Time trend has significant negative association in industry orientation. Results based on equation 6 shows that dummy for high income states has significant positive association with service sector and significant negative association with agriculture and manufacturing sectors.

(6.1)

(6.2)

(6.3)

(6.4)

(6.5)

In Table 7 the regression results (Pooled OLS) for the agriculture and its allied orientation of states and UT’s of India shows that high income states have significantly negatively affected by both agriculture and agriculture allied sector. The time trend seems to have significantly negatively affected industrial orientation of states and UT’s of India. Dummy for High income states and EAG states have significant negative effect on the manufacturing orientation. Though, we have removed multicollinearity in all specifications, as shown by Durbin Watson values, yet picture is not clear and hence we proceed with panel diagnostic tests in further analysis.

Finally, after panel diagnostic tests we find that fixed effect model is unbiased only for industry orientation and random effect models were found appropriate for agriculture and allied, agriculture, manufacturing and services orientations for states and UT’s of India for the period of 2004-05 to 2013-14. Now, it is crystal clear (Table 8) that any change in the share of agriculture and its allied sector, agriculture sector and even in service sector would have significant negative impact on per capita net state domestic product for states and UT’s of India. However, any change in share of manufacturing sector and industry sector significantly positively affected per capita NSDP. Change in population density has significant positive effects on the patterns of industrial sector. However, coefficient is insignificant yet positive for manufacturing orientation. These relationships suggest that most densely populated states can achieve economies of scale, resource endowments and scale of domestic demand easily and hence population density plays an important role in the patterns of industrial and manufacturing development.

Table 7. Augmented Chenery-Syrquin Structural Equations for States and UT’s of India, 2004-13

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Table 8. Estimation Results Using Fixed V/S Random effects, 2004-13

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

6. Conclusions

Chenery and other authors made a seminal contribution to the conceptualization of factors that affect structural change. Building on their conceptual framework, this paper first confirmed the strong association of economic structure with the level of income. The nature of relationship between per capita NSDP and agriculture and its allied sector is negative, while on the other hand relationship between per capita NSDP and industry or manufacturing sector are both positive.

Regression results of Pooled OLS for the agriculture and its allied orientation of states and UT’s of India shows that high income states have significantly negatively affected both agriculture and agriculture allied sector. The time trend seems to have significantly negatively affected industrial orientation of states and UT’s of India. Dummy for High income states and EAG states have significant negative effect on the manufacturing orientation.

Results of random effect model show that any increases in the shares of manufacturing sector and industry sectors have significant positive effect on economic growth (Income Coefficient), while the patterns of industrial sector has significant positive effects on population density (Size Coefficient). However, the coefficient of population density is insignificant yet positive for manufacturing orientation. These relationships suggest that most densely populated states can achieve economies of scale, resource endowments and scale of domestic demand easily and hence population density plays an important role in the patterns of industrial and manufacturing development.

Notes

1. Spurious regression means that the results obtained will suggest a statistically significant relationship between the variables in regression model when in fact all that is evidence of contemporaneous correlation rather than meaningful causal relation.

References

[1]	Bhatacharya, B.B. and A Mitra (1997): “Changing Composition of Employment in Tertiary Sector: A Cross Country Analysis”, Economic and Political Weekly, 21 March, Vol. 32, 529-34.
	In article
[2]	Blecker, R.A., 2009. External shocks, structural change and economic growth in Mexico, 1979-2007. World Development 37, 1274-1284.
	In article
[3]	Chenery, Hollis B., “Patterns of Industrial Growth,” American Economic Review, Vol. 50, No. 3 (September 1960), pp. 624-54.
	In article
[4]	Chenery, Hollis B. and Lance Taylor, “Development Patterns: Among Countries and Over Time,” Review of Economics and Statistics, Vol. 50, No. 4 (November 1968), pp. 391-416.
	In article
[5]	Chenery, Hollis B. and Moises Syrquin, Patterns of Development: 1950-1970 (New York: Oxford University Press for the World Bank, 1975).
	In article
[6]	Elhiraika, A. B. (2008). Promoting manufacturing to accelerate economic growth reduce volatility in Africa. Paper prepared at the African Economic Conference, jointly organized by the African Development Bank and UNECA in Tunis, Tunisia.
	In article
[7]	Kalirajan, K. (2004). “Economic reforms and the transmission of growth impulses across Indian states”, International Journal of Social Economics, 31 (5/6), 623-636.
	In article
[8]	Keesing, D.B. and Sherk, D.R> (1971), “Population Density in Patterns of Trade and Development”, American Economic Association, 61(5):956-961.
	In article
[9]	Kuznets, S. (1966), Modern Economic Growth: Rate, Structure and Spread, New Delhi, Oxford and IBH Publishing Co.
	In article
[10]	Libanio, G. (2006). Manufacturing industry and economic growth in Latin America: A Kaldorian approach. CEDEPLAR, Brazil: Federal University of Minas Gerais.
	In article
[11]	Linden, M. and T. Mahmood, (2007), Long run relationships between sector shares and economic growth: A panel data analysis of the schengen region. Economics and Business administration University of Joensuu, Finland Nachane, D.M., S.D. Sawant& C.V. Achuthan, (1989), Agriculture and industry: A study of selected linkages. Indian J. Agric. Econ., 44: 140-149.
	In article
[12]	Syrquin, Moshe and Hollis B. Chenery, “Patterns of Development, 1950 to 1983,” World Bank Discussion Papers, No. 41 (Washington, D.C.: World Bank, 1989)
	In article
[13]	Syrquin, Moshe, Lance Taylor, and Larry E. Westphal (eds.), Economic Structure and Performance: Essays in Honor of Hollis B. Chenery (Orlando: Academic Press, 1984).
	In article
[14]	Wang, S., & Li, D. (2010) “A empirical analysis on the relationship between service industry and economic growth”, Proceedings of 2010 International Conference on Industry Engineering and Management.
	In article