Madhya Pradesh: Does Agriculture Determine the State’s Growth Trajectory?

1. INTRODUCTION

This article makes an attempt to empirically analyse whether agricultural growth performance in the Indian state of Madhya Pradesh determines the state’s overall growth trajectory. The question of a state’s growth performance, in particular of agriculture, has often been the focus of various policy debates and has recently gained renewed attention. Despite recent growth momentum in agriculture, the annual growth performance of agriculture in Madhya Pradesh continues to be volatile, primarily on account of monsoon factors. In response, the state government’s initiatives on water management and extending irrigation facilities have shown modest results in the creation of irrigation potential and increase in net irrigated area. The state’s Economic Survey (Government of Madhya Pradesh, 2011) records that agriculture in the state remains traditional and continues to play a significant role in the state’s economic progress. At a time when state policy focuses centrally on industrial investment promotion and social engineering in rural areas through welfare schemes, the question of analysing agricultural growth performance regains its importance.

Historically, long-term trends of nearly three decades in Madhya Pradesh show that fluctuations in agricultural growth remarkably coincide with fluctuations in GSDP growth and have considerable impact on the state’s growth performance. In order to investigate such growth linkages, the agricultural terms of trade are introduced to capture the conventional agricultural supply curve and its responsiveness to price and economic incentives.

In analysing the growth performance of a state and agriculture in particular, the role of inter-sectoral terms of trade has been investigated less frequently. Ghosh (1988), Singh (1989), Misra and Hazell (1996a, 1996b), Raghavan (2004) and Deb (2005) among others, have explored the role and behaviour of the terms of trade using alternate data series and cross-sectional methods. At a state-economy level, terms of trade can be estimated at the sectoral level, which indicates the relative valuation of output of various sectors in the economy. This is a broad indicator of the supply side of the economy which can be used to analyse how price incentives, resources, production, investment decisions and policy initiatives have been instrumental in bringing a change in the structure of the economy. In what follows, Section 2 surveys a few studies on the state of agricultural and economic development in Madhya Pradesh, Section 3 describes the framework, methodology and estimation procedure, Section 4 discusses the findings and Section 5 concludes with a view on policy perspectives.

2. BACKGROUND AND PREVIOUS STUDIES

Madhya Pradesh has predominantly been an agrarian state. With large dependence of its population on the primary sector and diverse inter-regional differences, the state continues to face low levels of socio-economic development compared to other progressive states in India. In recent years, the annual growth performance of agriculture has shown major fluctuations with a growth rates ranging from a low of –2.44 per cent in 2007–08 to 10.87 per cent in 2009–10 and 1.18 per cent in 2010–11. During these periods per capita income has seen a modest rise by an average 4 per cent (Government of Madhya Pradesh, 2011). The sectoral contribution of agriculture in the state’s GSDP has gradually declined from about 25 per cent in 2004–05 to 20.53 per cent in 2010–11. Despite this declining trend, the percentage of workforce employed in the primary sector continues to be nearly 70 per cent, thereby making the growth performance of agriculture crucial for the socio-economic well-being of people engaged in this sector.

The issue of the state’s developmental roadmap and its policy divide on promoting industrial investment and agricultural modernisation has been debated for long. Ghosh (2005) critically analyses the structure of the state economy and highlights its widening duality. With large differences in the growth rates of the agricultural and non-agricultural sectors, the distributive gains and developmental progress in the state have remained limited. Pani (2007) investigates the performance of the state’s industrial sector after its bifurcation in 2001. The study argues that despite investment promotion, the early signs of industrial development were not promising as most indicators, such as, capital formation, employment and labour productivity, among others, failed to show any significant improvement during the last decade. On the agricultural front, Shankar (2005) reviews four decades of agricultural development in the state. The study argues that lack of regional or location-specific policies and unsustainable irrigation practices led to a near-stagnation of agricultural development thereby leading to technological and economic backwardness of many regions in the state.

In recent years, state policy has taken a renewed approach to overcome the variations in agricultural growth and also promoting industrial investment in the state. Initiatives of water harvesting and sustainable irrigation have resulted in modest improvements in expanding irrigation in monsoon-dependent areas and areas with water scarcity. In contrast, the promotion of investment through industrial summits and social engineering in the rural areas through several state-sponsored welfare schemes has remained a policy priority. Arguably, with the prevailing agricultural distress and a failure to achieve a sustainable agricultural and overall growth momentum, it is imperative that state policy be reorganised towards modernising and reforming agriculture and allied activities. This would greatly help in achieving a sustained livelihood for the large sections of the population engaged in these activities.

In order to make this assessment, growth linkages between agriculture and the overall state economy need to be explored. The following section develops an empirical framework by introducing agricultural or inter-sectoral terms of trade to analyse the supply response and cointegration analysis to explore the growth linkages.

3. MODEL SPECIFICATION

Conventionally, the state economy is represented by a broader sectoral classification, namely, primary, secondary and tertiary. These are further classified into 17 sub-sectors ranging from agriculture to services. For the present purpose, the state economy is aggregated into two major classifications, namely, agriculture (A) and non-agriculture (NA). The agricultural sector includes agriculture and allied activities, such as, forestry, logging and fishing. The non-agricultural sector includes mining and quarrying, and the entire secondary and tertiary sectors. ¹ , ² The data series are taken at constant and current prices from the State Domestic Product of India published by the Economic and Political Weekly Research Foundation (EPWRF), the Central Statistical Organisation (CSO) and the Reserve Bank of India (RBI).

Since data was available on different base-year series, a comparable back series of current and constant prices was constructed using conversion factors. Price deflators were constructed for all the series, and were used to arrive at the inter-sectoral terms of trade (ToT) as: _A/NAPrice IndexAPrice IndexNA

where A/NA represents the ToT for agriculture (A) vis-à-vis the non-agricultural sector (NA) and Price Index is the deflator series of the respective sectors. The concept of using deflators to measure TOT is akin to an ‘income’ ToT which indicates the relative valuation of agriculture to non-agriculture products, net of intermediate inputs. This concept also implicitly assumes production volume gains, as it takes into account the volume of agricultural and non-agricultural GSDP.

Theoretically, incorporating ToT as an explanatory variable gives us a variant of an agricultural supply response function. However, since a supply response function may also include other covariates, the ToT acts as the price determinant of the agricultural supply curve. Early attempts, such as, Bapna (1980), Singh (1989) and later Misra and Hazell (1996a, 1996b) among others, introduced simple versions of ToT models to capture the price and non-price determinants of agricultural supply. Desai and D’Souza (1999) pointed out some limitations in the models of Misra and Hazell (1996a) and indicated that their formulation does not adequately capture the price effect for agricultural supply. Alagh (2004, 2011) also formulated a ToT model by incorporating an acreage response function using all-India agricultural data. Deb (2005) conducted cointegration analysis between ToT and agricultural supply response and, among others, found that the ToT series are essentially non-stationary. Dholakia (2010) investigated the role of inter-sectoral ToT in leading to a growth momentum in agriculture in Gujarat. The study used structural breaks and piece-wise regressions to analyse different phases and varying behaviour of the ToT. Against this background, this article develops a framework to explore both the time series and structural properties of the ToT series.

In its simplest form, the basic time series properties are explored by fitting a linear time trend. The model is of the type: ToT_t(1)α₀α₁t

where (t) is time in years and (α₁) is the trend coefficient. However, this implicitly assumes stationarity of the time series and points to a deterministic trend in the series. The trend analysis is important from an analytical perspective as it has a significant bearing on ToT being an explanatory variable. It has been argued (Deb, 2005) that the presence of a unit root (hence a non-stationary process) could lead to a spurious relationship among variables in the regression, which could invalidate the results. Empirically, it is convenient to estimate the presence of a unit root and a simple model, akin to an AR(1) process, which can be specified as: ToT_t(2)α₀α₁ToT_t–1υ

where α₀ is a drift parameter and α₁ is the trend coefficient. A stationary process requires that |α₁| < 1. An alternate and more elaborate specification which controls for first-order serial correlation and includes a trend component can be written as: ToT_t(3)β₀β_jToT_t–jθtγ ToT_t–1u_t

where ∆(ToT) is (ToT_t – ToT_t–1) representing the differenced series of first order [I(1)] and the model includes an appropriate number of lagged terms (t < j). This is the usual Augmented Dickey-Fuller test (ADF) for checking the presence of a unit root with the null hypothesis, Ho: θ = γ = 0 using the F test if there is a trend and Ho: γ = 0 using a (t) test if there is no trend variable. Alternatively, the critical values of the tau (t = τ) statistic can also be used to evaluate the hypothesis. Both these models are used to identify the presence of a unit root in the series.

To further explore the linkages between ToT and the agricultural sector in the economy, this study investigates whether the ToT and the agricultural output series have any long-term relationship. This is estimated using cointegration, following which an error correction mechanism is incorporated. The underlying argument being that, while two series may exhibit random walk behaviour individually, they could well move together in the long run. Thus, if the agricultural output and ToT series were to be cointegrated, then it is not merely sufficient to use the differenced series for estimation purposes, as the information common between them is left out. An error correction model (ECM) would be suitable which will incorporate deviations from the long-term relationship as an explanatory variable. Similarly, the long-term relation between GSDP growth and agricultural growth can be explored. To begin with, in case of ToT, it can be stated that if agricultural output (Y) is estimated as a function of relative prices (ToT), then the equilibrium error in the long-term relationship can be expressed as:

e ^ t ≡ Y t − β ^ 0 − β ^ 1 T o T t

(3a)

In the event of both output and ToT being cointegrated, the error in the series is stationary, that is, I(0), which is further incorporated as an explanatory variable. The ECM using the differenced series of output and ToT is specified as (4) which may also include lagged values of the independent variable among other covariates.

Δ Y t = β 0 + β 1 Δ T o T t + β 2 e ^ t − 1

(4)

A similar procedure is followed to check for cointegration between GSDP and the agricultural growth series. Having explored cointegration, the empirical framework permits us to estimate the nature of causality between output growth, ToT and the growth series. However, given that causality cannot be estimated with certainty and remains an empirical difficulty, we can adopt the error correction mechanism to estimate the Granger causality or ‘precedence’ of growth momentum in agriculture and the overall state economy. Thus, one may expect a unidirectional causal relation between output growth and relative prices, but a bidirectional causal relation can also be explored. This is estimated using the Engel-Granger framework as: Y_t(5)β₀β_jY_t–jc_jToT_t–ju_t

where (Y) is agricultural output and ToT is agricultural terms of trade as earlier. The underlying argument being that, if past values of the ToT determine current values of (Y), then the ToT is said to Granger cause (Y). Alternatively, if changes in ToT precede changes in (Y), we can rule out Y being a cause of changes in ToT. Empirically, one can estimate the bidirectional causality using this framework and the magnitude of causality can be captured by ∑ c_j as the short-term effect and by ∑ c_j /(1 – ∑ β_j) as the long-term effect of ToT in this case. This model is used here to estimate the nature of causality between GSDP growth and agricultural growth.

Apart from time series investigation, some of the earlier versions of agricultural supply response models using ToT (Alagh, 2004, 2011; Misra and Hazell, 1996a, 1996b) were specified with agricultural output as a function of relative prices and a collection of non-price factors for a relatively short time period. Models such as:

Y = α 0 + α 1 T o T + α 2 A r e a + α 3 H Y L + υ

(6)

in which Area is the gross cropped area in hectares and HYL is the percentage of land under High Yielding Variety cultivation to account for non-price factors. Some other covariates (Z_it), such as, expected prices and lagged relation of output response to prices, have also been considered (see Alagh, 2004):

Y t = α 0 + α 1 P t − 1 + α 2 Y t − 1 + ∑ i α i Z i t + υ

(7)

The log version of this provides the price elasticity of supply (or acreage response) denoted by:

ln Y t = α 0 + α 1 ln P t − 1 + α 2 ln Y t − 1 + α 3 ln Z t + υ

(8)

where (α₁) and (α₃) are the elasticity estimates for price and non-price factors.

Equation (8) can be estimated using agricultural output or gross copped area as a function of lagged relative prices (ToT), lagged output or area in the previous period and controlling for other suitable non-price factors. However, time series data at the state level on several possible variables for non-price factors is limited and remains unavailable for long time periods. ³ Given the long-term trends and fluctuations of agricultural and overall non-agricultural prices, it is also worth examining whether such prices tend to move together. This has an interesting dimension as the movement of relative prices indicates economic linkages and resource flow in the economy and indicates supply-side decisions in the economy. One can thus examine non-agricultural sectoral prices (P_NA) as a function of agricultural prices (P_A) and understand the responsiveness of the non-agricultural sector to the changing environment of the agricultural sector.

Before we proceed to estimation, it is imperative to take into account aspects which may have a noticeable impact on relative prices. One of these is adverse supply shocks: droughts, floods, below-average rainfall, etc., are factors that negatively affect output and hence are likely to cause an upward pressure on agricultural prices. Empirically, adverse supply shocks have been controlled by introducing dummy variables by identifying years of such shocks to account for large negative deviations in output. ⁴ The second issue is to identify structural breaks in the long-term series of agricultural ToT. This is achieved by using the endogenous method following Bai and Perron (1998, 2003), Zeileis et al. (2002) and Wang (2006). The model used is the basic trend function allowing for a break in the level (intercept), thus indicating different regimes of ToT. The following section discusses the results of the models and the growth performance.

To analyse the growth performance and linkages between the agricultural sector and the overall economy of the state of Madhya Pradesh, the study begins with a phase-wise analysis (Table 1). The phase-wise performance of agriculture along with trends in the ToT reveals some interesting facts. Agricultural growth in the state has been highly volatile, as captured by the coefficient of variation during the phase-wise five-year periods. Variations in agricultural growth rates have been particularly high when relative prices were fluctuating. Given such large fluctuations, the short period averages do not entirely capture the growth trajectory of the sector. However, a graphical analysis of annual agricultural and GSDP growth rates clearly brings out the fluctuations (Figure 1).

It is fairly evident that fluctuations in agricultural growth entirely coincide with GSDP fluctuations (Figure 1). Agricultural growth at a regional level is assumed to show considerable fluctuations on account of several natural and state-specific factors. On this premise, the role of relative prices and agricultural supply response gains a primal focus.

Figure 2 shows the trend of inter-sectoral ToT for agriculture vis-à-vis the non-agricultural sector for the past three decades. Broadly, the agricultural ToT for the state shows a secular upward trend without major cyclic fluctuations. The trend suggests turning points over some years where the ToT has changed from negative to positive in favour of agriculture.

Thus, to empirically investigate, consider first the responsiveness of non-agricultural prices to changes in agricultural prices. As mentioned, this is indicative of the fact that resources would move as per favourable prices and hence would determine investment and profit potential in both sectors. The estimated equation of non-agricultural prices as a function of agricultural prices after controlling for trending variables and adverse supply shock is as follows: P_NAt(9)P_AtP_NAt–1tD_t^#2

The coefficient on ln P_A shows a positive elasticity of 0.16 with respect to current non-agricultural prices. However, this result may not hold given the presence of serial correlation, structural differences in ToT over time and other determinants that may not have been included herein. However, the Breusch-Godfrey LM test statistic χ²(p) gives a value of 1.200 and the corresponding (p) value (Prob > χ²) as 0.273 thereby indicating no serial correlation. Furthermore, in order to see the trend and structural properties more explicitly, the empirical findings of the following models need to be considered.

OPEN IN VIEWER

Figure 1

Source: Compiled from EPWRF (2009) and Government of Madhya Pradesh (2011).

OPEN IN VIEWER

Figure 2

Source: Compiled from EPWRF (2009) and Government of Madhya Pradesh (2011).

The result of the first model which estimates a basic trend in the long-term ToT is: ToT_t(10)t²

This indicates that the long-term agricultural ToT have shown a positive and consistent improvement over time. However, given that such a series may be non-stationary, the results of models (2) and (3) are used to check for stochastic properties while controlling for serial correlation and a time trend: ToT_t(11)ToT_t–1ToT_t–1ttα₁₌₀²

The test statistic tau (t = τ) for n < 50, at the 0.01 and 0.05 levels of significance inclusive of a trend and constant terms, are 4.15 and 3.50, respectively. The computed (t = τ) values for the model are less than the critical values which confirms that the ToT series is non-stationary. This thus indicates that taking the ToT as an explanatory variable may well lead to a spurious relation with agricultural output. Similarly, before modelling agriculture supply, the stochastic properties of the output series are analysed. Equation (12) gives the following result: Agri_t(12)Agri_t–1ttα₁₌₀²

As the computed (t = τ) value is marginally less than the critical value at the 5 per cent level of significance, the result shows that the output series is non-stationary. This leads to the question of whether the output and ToT series have any long-term relation. If the output and ToT series were to be cointegrated, then a linear combination of the two must be stationary.

Consider the basic model (13) of output and ToT using log variables: Agri_t(13)ToT_ttα₁₌₀²

In the event of cointegration of the two series, the equilibrium error obtained from this, using Equation (3a) must be stationary. Checking for a unit root in the estimated error the following equation gives the result:

Δ e ^ t = – 0. 5991 e ^ t − 1

(13a)

The computed (t = τ) statistic (for α = 0 without the intercept) is greater than the critical value which rejects the null hypothesis of a unit root. Thus, the error of a linear combination of the two series is stationary, which indicates that both series are cointegrated. One can further use the error information to model the long-run relationship between the two series after controlling for other factors. Since both output and the ToT series are I(1), they are only stationary in their first differences.

The study, therefore, formulates an ECM by replacing output and the ToT series with their first differences, includes a lagged ToT term to capture the relative price response and the lagged error term obtained from the cointegration equation. Thus, the ECM is formulated as: Agri_t(14)ToT_tToT_t–1–

The sign on the error coefficient is negative which shows that the model is stable, as a negative correction takes place to restore equilibrium in agricultural output. The error value is statistically insignificant which indicates that the correction happens in the current period only to restore equilibrium.

Since output and relative prices follow a long-term relation it is also worth exploring whether growth of the agricultural sector and growth of the state economy have any long-term relation. Using the same ADF framework, first the stationary properties of the agricultural and GSDP growth series are analysed. Equations (15) and (16) are estimated to check for unit roots in both series: Gr.GSDP_t(15)Gr.GSDP_t–1Gr.GSDP_t–1ttα₁₌₀² Gr. Agri_t(16)Gr. Agri_t–1Gr. Agri_t–1ttα₁₌₀²

Both results have (t = τ) greater than the critical value and hence we can infer that both series are stationary. Alternatively, consider a variable {S_t} which is the difference of the annual growth rate of GSDP and agriculture. Thus, S_{t = (Gr. GSDP – Gr. Agri.). Since both series may have a long-run relation, a linear combination of them must be stationary in the event of cointegration. Using this information, the first step is to make the cointegrating equation between GSDP and agricultural growth by formulating the basic model as: Gr.GSDPt(17)Gr.Agrit²}

As earlier, the estimated equation for stationarity of the equilibrium error obtained from (17) is:

Δ e ^ t = – 0. 858 0 e ^ t − 1

(17a)

Alternatively, one can also check for cointegration using the variable of difference in growth rates {S_t}. Similar to the earlier error equation, one can estimate whether the variable {S_t} is stationary. The estimated equation is: S_t(17b)S_t–1tα₁₌₀²

Both results confirm that GSDP growth and the agricultural growth series are cointegrated and thus have an equilibrium long-term relationship. One can, therefore, model this relation using an ECM by including relevant lagged variables as other covariates. The ECM is formulated as:

(18)

The results show that a 1 per cent point change in the agricultural growth rate leads to a positive change of approximately 0.43 per cent in overall GSDP growth. The equilibrium error coefficient is negative and statistically significant, which explains a correction of nearly 50 per cent in GSDP growth as it deviates from its equilibrium long-run value. Using the ECM, we can estimate the short-term and long-term impacts of variation in agricultural growth on GSDP. Since there is a two-period lag, the short-term impact is given by ∑ c_j = 0.582, which explains approximately half the variation in GSDP growth. The long-term impact is given by ∑ c_j /(1 – ∑ β_j) which for j = 2 approximates to [0.582/1–(–0.402)] = 0.41. Thus, in the long term, variations in agricultural growth explain nearly 40 per cent of the variation in GSDP growth. These results suggest a considerable impact of agricultural growth on the overall growth of the state economy both in the short and long run, thereby clearly establishing it as a significant determinant of the state’s growth trajectory.

Having explored the growth linkages, we now consider the structural properties of the ToT. This is investigated using an algorithm for estimating structural breaks in the long-term trend of agricultural ToT. For the present purpose, a simple linear trend model of the type ln ToT = α + β(t) + υ is adopted with breaks allowed in the intercept. Thus, structural breaks are referred to as breaks in the level over time with an unknown number of possible breaks in the series. The algorithm is based on the method proposed by Bai and Perron (1998, 2003) and Zeileis et al. (2002). ⁵

The result for break dates estimated in the level of ToT is as follows:

Segment

m = 1 1997

m = 2 1986 1997

m = 3 1986 1997 2004

m = 4 1986 1992 1998 2004

The result shows two break dates, namely, 1986−87 and 1997−98, based upon the minimum BIC criteria. Graphically, these are plausible break dates where the ToT have shown a clear shift to a different trajectory. There are, therefore, three regimes in which ToT trends can be analysed: 1986−87 marks the beginning of a consistent upward trend in the ToT and subsequently 1997−98 also shows a minor upward improvement in the trajectory. As these periods are structurally different, break dates are used to calculate the phase-wise trend growth rates of GSDP and the agricultural sector (Table 2).

The trend in growth rates for agriculture and GSDP present a fairly accurate record of growth performance in recent years. The trend for agriculture is negative while the GSDP growth rate is under 2 per cent in the initial period. Subsequent periods show a considerable improvement in GSDP growth as agriculture is able to register a consistent growth rate. It is interesting to note that the agricultural growth rate had risen during periods of favourable ToT and in particular GSDP shows a higher and consistent trend rate during rising agricultural ToT. However, the trends and growth linkages may have a much broader relationship, as output responds to several price and non-price factors. Theoretically, it is based on the fact that relative prices would have an a priori positive relationship with the level of output. This further suggests that one may postulate a functional relationship between the level of output and ToT to determine the responsiveness of changing ToT on the level of output or growth. The underlying economic intuition is that as relative prices improve for the sector in question, this translates into economic incentives for producers to enhance production which in turn indicates higher growth prospects. Thus, the relative price elasticity of supply can be estimated to understand the responsiveness of movements in agricultural prices to agricultural supply and also the output response of the overall state economy. Since agricultural output might have a lagged response to rising prices, the study incorporates lagged relative prices to capture the time effect in the adjustment process of agricultural supply and prices. Table 3 presents the elasticity estimates.

OPEN IN VIEWER

Table 2

Trend Growth Rates (%) for GSDP and Agriculture (2004–05 constant prices)

Years	GSDP Growth	Agriculture and Allied Sectors Growth	Agricultural Terms of Trade
1980−81 to 1986−87	1.98	–2.32	Falling
1986−87 to 1997−98	5.24	3.18	Fluctuating
1997−98 to 2010−11	5.34	2.94	Rising

Source: Compiled from EPWRF (2009) and Government of Madhya Pradesh (2011).

OPEN IN VIEWER

Table 3

Relative Price Elasticity of Agricultural Supply

	Dep. Var.	Constant	Coefficients
1.	ln Agrit	4.72	0.445 ln(PA/PNA)t–1	+ 0.687 ln Agrit–1	– 0.184 D
			(0.180)∗	(0.105)∗	(0.036)∗
				R2 = 0.87, N = 31

Note: Values in parentheses are standard errors. ∗ indicates value significant at 1 per cent level.

The estimation is done with covariates of lagged relative prices, lagged agricultural growth and supply shocks (such as droughts). The coefficient on lagged relative price is positive and significant indicating a positive and conventional relation of the supply curve. The coefficient indicates a lagged supply elasticity of 0.44, or 44 per cent of the output for a unit change in relative prices. This suggests that output responds to rising prices in agriculture, as this shows a conventional rising supply curve. The dummy variable shows a consistent result as it is negative and significant, thereby indicating a larger negative deviation in agricultural output on account of adverse supply shocks. These results confirm the significant impact of agricultural growth performance on the overall growth trajectory of the state. As agricultural growth shows considerable responsiveness to price incentives, it indicates that agricultural output can be augmented through economic incentives, investments and technological advancement.

This article empirically analyses whether agricultural growth performance in the Indian state of Madhya Pradesh determines the overall growth trajectory of the state economy. Long-term trends over nearly three decades show that fluctuations in agricultural growth clearly coincide with fluctuations in GSDP growth. To analyse this, the dimension of agricultural ToT is introduced to capture the agricultural supply response to relative prices and economic incentives. The findings show that the agricultural sector in Madhya Pradesh demonstrates a credible growth performance under a favourable ToT regime. This is further supported by positive and significant relative price elasticities of agricultural output, indicating that agriculture in the state is responsive to prices and economic incentives. The results of the endogenous method for identifying break dates show two breaks, namely, 1986–87 and 1997–98, in the ToT series, thereby making three different regimes, namely, 1980–87, 1987–97 and 1997–2011. These three regimes clearly point to different growth trajectories of agriculture in the state with a move to a higher trajectory under rising agricultural ToT.

Empirically, agricultural growth and relative prices show cointegration based on which an error correction mechanism is used to model their long-term behaviour. Similarly, agricultural growth and GSDP growth also show cointegration which is further explored using the Engel-Granger method for estimating causality. The findings show that in the short term, fluctuations in agricultural growth explain nearly half the fluctuations in the state economy’s growth, whereas the long-term impact is close to 40 per cent. With favourable ToT and positive growth linkages, the findings contain corroborative evidence that agricultural growth remains a key determinant of the state’s growth trajectory.

This reflects on the policy divide between investment promotion and social engineering in the state. As the state gathers momentum to achieve a high growth trajectory through industrial investments, the solution may well lie in modernising agriculture and developing the agriculture–industry linkage. Given considerable agricultural supply response, investments, technological improvements and remunerative prices in the primary sector, the state could continue on the growth momentum and ensure a sustainable high growth trajectory.