ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Impact of Declining Groundwater Levels on Acreage Allocation in Haryana

This study attempts to examine the impact that a declining groundwater table has on the farmer's crop response function in Haryana. The paper first summarises the most important causes for depletion of groundwater in Punjab and Haryana. It then reports a regional study of Haryana using the Nerlovian supply response function. Supply response is estimated for two water-intensive crops (rice and sugar cane) and two less water-demanding crops (wheat and bajra). The Arellano-Bond Dynamic Panel estimator is used in the study. The regression results confirm the hypothesis that farmers are recognising and responding to declining depths of water tables. Farmers are reducing the acreage devoted to rice and sugar cane, the two most water-intensive crops but not for the less water-intensive crops.

Impact of Declining Groundwater Levels on Acreage Allocation in Haryana

This study attempts to examine the impact that a declining groundwater table has on the farmer’s crop response function in Haryana. The paper first summarises the most important causes for depletion of groundwater in Punjab and Haryana. It then reports a regional study of Haryana using the Nerlovian supply response function. Supply response is estimated for two water-intensive crops (rice and sugar cane) and two less water-demanding crops (wheat and bajra). The Arellano-Bond Dynamic Panel estimator is used in the study. The regression results confirm the hypothesis that farmers are recognising and responding to declining depths of water tables. Farmers are reducing the acreage devoted to rice and sugar cane, the two most water-intensive crops but not for the less water-intensive crops.


ver the years there has been overexploitation of groundwater, which has been used to meet the increasing demand for water. This has resulted in declining water table in various parts of the country. It is believed that one of the major causes of the decline in groundwater tables is the introduction of water-intensive crops such as paddy and sugar cane into the cropping pattern in certain regions. Yet, the relationship between groundwater depth and changes in the cropping pattern has not been investigated adequately in the literature so far. Furthermore, there is no analysis of how farmers in turn have adapted their behaviour in terms of decisions relating to acreage allocation in response to the declining water table. It is this gap that this study bridges.

The focus is on Haryana, where the depletion of groundwater resources could threaten the long- term sustainability of irrigated agriculture on which the state depends. The study is divided into four sections. Section I examines trends in cropping patterns and groundwater situation in various regions of Haryana. Section II sets out the data and methodology used. Section III provides an analysis of the results obtained through various supply response based econometric exercises. Section IV concludes and draws implications for policy.

The analysis is undertaken at the regional level (to minimise aggregation bias) and focuses not only on rice, wheat and sugar cane, but also three dryland crops – bajra, maize and jowar. These crops account for almost 70 per cent of the area under cultivation in the state. The study considers all regions at the outset but focuses specifically on the analysis of only those regions facing the problem of groundwater depletion. It analyses the impact of falling water levels on cultivation of crops using a modification of Nerlove’s supply response model. A rational farmer should respond to hydrologic conditions that are outside his control by planting more or less of a certain crop. The way farmers react to this certainly has a crucial impact on the availability of resources in future. The hypothesis to test, therefore, is that acreage of water-intensive crops and depth of groundwater should be negatively related. Prices are also crucial in a supply response model. A positive relation should exist between area cultivated and price of the crop under consideration.

The primary cause of decline in water table is said to be the introduction of water-hungry crops in Haryana. Since there is close correspondence between expansion of area under paddy in a wheat rotation (which is popularly practised in Haryana) and incidence of water depletion, it is strongly believed by many that paddy-wheat rotation is the main cause of the problem. Numerous studies, for example, Chand (1996), Chand and Haque (1997), Bathla (1996), Kataki, Hobbs and Adhikary (2001) and Duxbury (2001) hold that both in Punjab and Haryana the paddy-wheat rotation is unsustainable and is lowering the water table. The increased popularity of these crops may be attributed in part to the government’s price policy. It is important to note here that initially, at least, providing remunerative prices for rice and wheat was important not only for ensuring food security, but they also helped in draining out the excess water brought about due to unlined canals and lack of adequate drainage in the state. But cultivation of these crops over the years has brought about a massive decline in the water level. Therefore, the crops that were earlier encouraged have now become a cause of a depleting resource base. A contrary view on the cause of depleting water tables is provided by a study conducted by the Punjab Agricultural University, Ludhiana [Prihar et al 1990]. The study suggests that the paddy crop is not responsible for resource depletion and it provides estimates of actual water used by important crops net of seepage, i e, part of irrigation water that percolates back to water table.

Pricing of electricity is another reason responsible for largescale mining of groundwater. Due to a variety of reasons, but most importantly, due to administrative simplicity, many states in the north, notably Uttar Pradesh, Punjab and Haryana operate on a flat rate system.1 This pricing structure is responsible for wastage of both power and groundwater resources. Since the flat power charges do not vary with the hours of pumping, farmers have no incentive to economise on the use of electricity.

I Descriptive Statistics on Agriculture and Irrigation

Twin Problems of Depletion and Waterlogging

Estimates of groundwater depth in the state show that the groundwater level is generally high in the southern parts and low in the north and north-east, which is a hilly tract. During the pre-monsoon period, it ranges from 5m to 21m below ground level (bgl). Deepest water levels are observed (21m bgl) in the south-western part of the state especially in Mahendargarh district. Shallow water level conditions exist in Rohtak and Jind districts, i e, central parts of the state and in the districts of Bhiwani, Sirsa and Hissar (Table 1).

What is often not recognised is that the groundwater problem in Haryana has two dimensions. The first is that of rising groundwater table in the areas with low quality aquifers, leading to secondary salinisation and waterlogging.2 The second is that of declining water tables due to over-pumping of groundwater in fresh water quality aquifer zones. The districts of Haryana are accordingly classified into these two types.

This is amply demonstrated in Table 1 which provides a comparison of water levels from 1980 to 2002. Water levels have declined in Faridabad, Gurgaon, Mahendargarh, Panipat and Yamunanagar region. Among these Gurgaon, Panipat, Mahendargarh and Yamunanagar are the regions where the water table has fallen by more than three metres during this period. The depth has nearly doubled in Gurgaon and Panipat. Interestingly, the water table has declined both in the regions where the water table was high (such as Panipat) as well as those where the water table was deep (such as Mahendargarh). In the latter case the water table declined from 14m bgl to 21m bgl. In contrast, in the regions of Bhiwani, Jind, Hissar, Rohtak and Sirsa the water table has risen by nearly 5 metres in 20 years.

To capture the severity of depletion, it is common to categorise areas as being “dark”, “grey”, or “white” (Table 2). A block is characterised as “white” if the rate of groundwater exploitation is below 65 per cent, “grey” if the rate of exploitation happens to be in the range of 65 per cent to 85 per cent and “dark” if the groundwater use is above 85 per cent of its utilisable recharge. Of 106 hydrogeological blocks in Haryana, 44 per cent of them are categorised as dark zones. Most of these lie in zone I. Karnal, Kurukshetra, Mahendargarh and Rewari have all their blocks lying in the dark category. In fact, for some of these the utilisation ratio exceeds even 100 per cent of recharge – Kurukshetra (178 per cent), Karnal (132 per cent) and Mahendargarh (130 per cent) [Dhawan 1995]. The situation is also dismal in the districts of Ambala, Gurgaon, Panipat and Yamunanagar. Out of the rest, 17 blocks (16 per cent) fall under the grey zone category, where groundwater exploitation is almost equivalent to recharge. This leaves only 39 per cent of the area categorised as white zone, the majority of which are in zone II. For instance, only Hissar and Rohtak have none of their blocks in the dark category. Further, these include the areas where groundwater is saline and therefore not suitable for cultivation of most crops.

Trends in Cropping Pattern

For the state as whole, (Tables 3a and 3b) wheat covered the largest share of the gross cropped area in both 1981-83 and 19982000. The share of bajra dropped from 15 per cent to 10 per cent while the share of rice doubled from 9 per cent to 18 per cent during the same period. Sugar cane and jowar occupied a much lower share in both the years (almost 2 per cent). Similarly, both maize and barley had an extremely low share in 1980s, which further dropped to being almost negligible in total cropped area in the late 1990s. Barley was grown in Gurgaon and Mahendargarh in 1980s. Again by the 1990s one observes a massive decline in cultivation of this crop in all regions of the state.

More interesting trends are seen on comparing the cropping pattern across the regions. This is the advantage of conducting a regional study. Over time rice has substituted for bajra. In 198183, Panipat and Yamunanagar (both regions facing the problem of water depletion) had substantial part of area under paddy, but had very low share of area under bajra. All other regions (except Sirsa, which does not depict this substitution pattern) had more of area under paddy and very little under bajra. This pattern was also seen in many regions in 1998-2000 (Panipat, Yamunanagar, Gurgaon and Jind). Thus it can be clearly seen that rice, a highly water-driven crop had been cultivated more and more over the years and has replaced bajra, jowar and maize, all three being dryland crops. Farmers in a region like Mahendargarh (which, earlier, had no area under rice as it has the deepest water level in Haryana) had also started cultivating this crop. It is interesting that the expansion in rice acreage at the expense of coarse cereals had occurred not only in the water depleting areas, but also those in which water tables are rising. Although sugar cane does not account for much more than 2 per cent of the cultivated area in the state as a whole, its share is increasing slightly in Yamunanagar and Faridabad and has decreased in nearly all other regions.

Thus the increase in the relative area under rice cultivation is consistent with declining groundwater tables in three regions – Faridabad, Panipat and Yamunanagar. It is only in the latter two regions, however, that rice cultivation accounts for over a quarter of the cropped area. The explanation for declining water tables in Gurgaon and Mahendargarh lies elsewhere, for there is virtually no rice cultivated in these districts. Similarly, prima facie,

Table 1: Depth of Water Table (m)

Region 1980 2002 Absolute Change

(1) (2) (3) (4) = (2) – (3)

Zone I: Water Depleting Areas Faridabad 6.5 8.2 -1.7 Gurgaon 6.7 13.2 -6.5 Mahendargarh 13.9 21.0 -7.1 Panipat 5.0 10.4 -5.3 Yamunanagar 7.9 11.3 -3.3 Zone I 8 12.8 -4.8

Zone II: Waterlogged Areas Bhiwani 20.5 16.5 4.0 Hissar 11.6 8.6 3.0 Jind 9.6 8.0 1.7 Rohtak 5.7 5.3 0.3 Sirsa 21.7 6.9 14.8 Zone II 13.8 9.1 4.8

Source: Groundwater Cell, Department of Agriculture, Haryana.

Table 2: Number of Dark, Grey and White Blocks on the Basisof Groundwater Assessment as on April 1, 1997

Number of Blocks

Dark Grey White Total

I Faridabad 1 2 2 5 Gurgaon 5 1 39 Mahendargarh 10 0 0 10 Panipat 10 4 3 17 Yamunanagar 11 4 6 21

II Bhiwani 3 1 48 Hissar 2 2 812 Jind 2057 Rohtak 2 2 610 Sirsa 1157 Haryana 47 17 42 106

Source: Groundwater Cell, Department of Agriculture, Haryana.

Economic and Political Weekly June 30, 2007

sugar cane can explain the decline in groundwater levels in at most 2 per cent of the states’ area under cultivation. Note also that as was the case in the regions that are now facing resource depletion, the expansion in rice acreage in zone II, may prove to be a way of solving the problems of waterlogging. The difference is that while much of the water in zone I is fresh, that in zone II is saline.

This regional analysis of cropping pattern also gives a basis for selecting the regions for which econometric exercise is carried out (Section III). Wheat is grown in all regions of zone I and II. Therefore for wheat, results are reported for zone I (the waterdepleting zone) and all regions of Haryana. Rice is grown largely in Panipat and Yamunanagar of zone I. Sugar cane is grown in Yamunanagar, Panipat and Faridabad of zone I. Accordingly, for rice and sugar cane the results are provided only for these regions. Amongst the dryland crop, bajra is chosen as maize and jowar have very low share of area.

II Data and Methodology


The study analyses available secondary data for all the 19 districts of Haryana from 1980 to 2000. The data on area, production and yield of crops is collected from the Statistical Abstract of Haryana. The data on farm harvest prices are collected from the Farm Harvest Prices of Principal Crops in India (from 1981 to 1999) and on minimum support price (MSP) from the reports of the Commission for Agricultural Costs and Prices (CACP). The data on groundwater was not readily available as it is not published in any source. The Groundwater Cell, department of agriculture, Haryana, records depth of groundwater (both pre- and post-monsoon) in all districts of the state.3 To gather this data the department of agriculture has been monitoring a network of 2021 observation wells distributed all over Haryana.

To bring the data to a consistent format and to enable comparison over the 20-year period some modifications were made to this data. There is a substantial regional heterogeneity in the state which necessitates the districtwise study of Haryana. However, over time district boundaries have changed, with larger districts frequently subdivided to create new ones. To maintain consistency in the data series, instead of 19 districts, the study concentrates on 10 regions each of which consists of one or more districts. These are Bhiwani, Faridabad, Gurgaon, Hissar (consisting of districts of Fatehabad and Hissar), Jind, Mahendargarh (consisting of districts of Rewari and Mahendargarh), Panipat (consisting of districts of Panipat, Karnal and Sonipat), Sirsa, Rohtak (consisting of districts of Rohtak and Jhajjar), and Yamunanagar (consisting of districts of Yamunanagar, Kurukshetra, Panchkula, Kaithal and Ambala).


In this study, the supply response function based on the Nerlovian Model will be estimated. Economic research on how farmers react to prices and other incentives was advanced by Marc Nerlove’s seminal work of 1958. The Nerlove model hypothesise farmer’s reaction based on price expectations and area (or production) adjustments to examine the supply response. This model states that output is a function of expected price, area (output) adjustment, and some exogenous variables. The standard structural form of Nerlove model consists of the subsequent three equations: A*t = a0 + a1P*t+a2 Zt + ut ...(1) P* =P*t-1 + β (Pt-1–P* ...(2)

tt-1) At=At-1 + γ (A* ...(3)

t–At-1) where, At is the actual area under cultivation at time t, A* is the

t desired area under cultivation at time t, Pt is the actual price at

time t, P* is the expected price at time t, Zt represents other

t exogenous factors affecting supply at time t, and β and γ are termed as the expectation and adjustment coefficients respectively. Equation (1) describes the relation between desired acreage and expected price. Equations (2) and (3) characterise simple adaptive expectations and partial adjustment mechanism linking desired price and area to observable price and area values. In equation (2), Nerlove postulates that this adjustment can be expressed as a fraction of the difference between last period’s actual and expected normal prices. In equation (3), Nerlove expresses actual output as adjusted by some fraction (γ, the area adjustment coefficient) of the variation between long-run equilibrium output (desired) and actual output in the previous period. The above equations contain variables that are expected (or desired) and are not observable. To estimate the observed variables a reduced form equation is derived after few manipulations from the above three equations. (1 – β+1– γ)At-1 – (1–β)(1–γ)At-2

At = α0βγ +α1βγPt-1 +

+ α2γDt – α2 (1–β)γ Dt-1 + γ [Ut –(1–β)Ut-1] (4) where, At , A t-1 and A t-2 are the acreage under a particular crop

Table 3a: Cropped Area as a Percentageof Gross Cropped Area in 1981-83

(Three-year averages)

Regions Sugar Cane Wheat Rice Jowar Bajra Maize Barley

I Faridabad 2.7 45.0 1.3 6.3 15.6 1.1 0.6 Gurgaon 0.3 35.6 0.2 5.1 21.5 0.0 7.7 Mahendargarh 0.0 18.8 0.0 1.1 37.7 0.0 6.0 Panipat 4.9 45.6 23.0 3.1 4.0 1.7 0.6 Yamunanagar 5.4 40.9 27.5 0.3 1.5 4.5 0.6

II Bhiwani 0.5 9.0 0.0 0.8 36.2 0.0 0.6 Hissar 0.7 22.6 2.6 0.3 12.8 0.2 0.8 Jind 3.1 33.4 8.5 2.3 18.1 0.2 1.1 Rohtak 5.1 31.2 0.8 10.4 18.6 0.0 2.3 Sirsa 0.0 22.9 3.7 0.1 2.9 0.1 1.2 Haryana 2.5 30.2 9.2 2.3 14.7 1.1 1.9

Source: Statistical Abstracts of Haryana.

Table 3b: Cropped Area as a Percentageof Gross Cropped Area in 1998-2000

(Three-year averages)

Regions Sugar Cane Wheat Rice Jowar Bajra Maize Barley

I Faridabad 3.1 49.0 10.3 9.0 6.6 0.2 0.6 Gurgaon 0.2 43.8 3.1 6.9 19.2 0.0 1.1 Mahendargarh 0.0 20.5 0.2 0.4 31.8 0.0 0.6 Panipat 2.7 43.7 35.8 1.6 1.0 0.1 0.0 Yamunanagar 6.8 38.8 36.7 0.0 0.6 1.8 0.0

II Bhiwani 0.2 16.5 0.8 0.7 28.9 0.0 0.6 Hissar 0.5 33.7 10.0 0.0 7.4 0.0 1.0 Jind 1.5 41.2 22.7 0.6 8.2 0.1 0.3 Rohtak 3.5 41.3 8.8 13.2 11.1 0.1 1.0 Sirsa 0.0 35.4 6.1 0.0 0.6 0.0 1.3 Haryana 2.2 35.5 16.7 2.0 9.6 0.4 0.6

Source: Statistical Abstracts of Haryana.

in period t, t-1 and t-2 respectively, Pt-1 is the relative price of the crop in period t-1, Dt and Dt-1 are the depth of water table in period t and t-1, respectively.

Before such an equation can be estimated, several issues need to be clarified. The first is what price series should be used on the right hand side. In this study both farm harvest prices and minimum support prices expressed in relative terms have been used. In the Indian context given that government announced minimum support prices virtually determine market prices, as an alternative, the model is estimated using MSP.

The second has to do with what variable to use to represent farm supply response on the left hand side. Output is measured either in terms of volume produced or marketed or yield per unit area. However, Askari and Cummings assert that the relationship between expected prices and farmers’ supply response decision is best expressed in terms of harvested acreage because this is how farmers translate their price expectations into actions. Accordingly, in this study we use acreage as a measure of farm supply response. More specifically, the ratio of crop acreage to gross-cropped area is used as the dependent variable in the regressions that are performed.

Besides, identifying what output and price measure to use, a decision needs to be taken on what variable to use as ‘Z’. Various studies have used proxies for technology, institutional factors, irrigation, etc. Depth of groundwater is the exogenous variable ‘Z’ in the present study. As noted earlier the coverage of highyielding variety seeds and irrigation is nearly 100 per cent in this state, throughout the period under study, making their inclusion in Z not meaningful. For kharif crops, i e, rice and bajra depth of groundwater (D) in the pre-monsoon period is used, while, for the rabi crops depth for the post-monsoon period is more relevant for farmers. For sugar cane which is an annual crop average depth of groundwater during a year is used.

The effect of rainfall is also built in this shift variable (Z or D in our study); therefore, it is not considered separately in the supply response equation. The presence of both rainfall and depth may lead to multicollinearity as depth will indeed be a function of rainfall. The correlation between depth and rainfall was found to be negative and significant. The correlation coefficient was -0.3. Risk is not considered in the present study, as it is not important in affecting farmers’ supply response in Haryana. Risk may arise due to variation in either price or yield. The existence of MSP reduces the price risk for the farmers, while the yield risk is also not important as much of cultivated area in Haryana is irrigated.

Thus, a crop-specific model, rather than a model of aggregate supply response is formulated in this study. An aggregate supply response model is inappropriate given the objectives of the study. Further, almost the entire cultivable area is already tilled in Haryana. This study makes use of adaptive price expectation, which is one simple way of formulating expectations

Estimation Technique

The estimation technique employed here takes explicitly into account the cross section and time series nature of the database. Moreover, pooling of data across all the regions will also increase the degrees of freedom. Dynamic model estimation is performed because of presence of a lagged dependent variable on the right hand side. Including lags of dependent variable in the right hand side of the equation induces a correlation between the error term and lagged dependent variable. This adds an element of complexity as in this case the ordinary least square estimator is inconsistent [Hsiao 2003]. A normal technique for dealing with variables that are correlated with the error term is to instrument them. There are a number of ways of doing so, the best known are Anderson-Hsiao and Arellano-Bond approaches. Anderson – Hsiao estimator may be consistent but is not efficient because it does not take into account all available moment restrictions.

Arellano-Bond (1991) argue that a more efficient estimator results from the use of additional instruments whose validity is based on orthogonality between lagged values of the dependent variables and errors [Li et al 2003]. It is appropriate for both time-series and cross section time-series (panel) regressions. One advantage with this method is that it takes account of first-order serial correlation of the error term and provides unbiased as well as efficient estimates. Therefore, in this study this estimator is used, perhaps, for the first time, for an agricultural supply response model. Consider a fixed effects dynamic model of the form:


yi,t = γ yi,t-1 +uit Taking first differences, (yit–yit-1)= γ (yit-1–yit-2)+(uit–uit-1) i=1,...,N t=1,...,T (6) Now we need to instrument, Δyit-1 = (yit-1 – yit-2), which is still clearly correlated with the error (uit – uit-1). The second lag of the level, yit-2, and the first difference of this second lag, Δyit-2 = (yit-2 – yit-3), are possible instruments, since they are both correlated with (yit-1 – yit-2) but are uncorrelated with (uit – uit-1), as long as the uit themselves are not serially correlated. Arellano and Bond suggest that one should use yit-2 rather than Δyit-2 as for the latter the variances are likely to be very large for certain ranges of autocorrelation (γ). Take (6) above, the firstdifferenced simple AR (1) model with no regressors. At t = 3, the first period we observe the relationship in (6), (yi3 – yi2) = γ(yi2 – yi1) + (ui3 – ui2) (7) yi1is a valid instrument for Δyi2, since these are highly correlated, and yi1is not correlated with (ui3 – ui2) unless the uit are serially correlated. At t=4, (yi4 – yi3) = γ(yi3 – yi2) + (ui4 – ui3) (8) Here yi2 and yi1 are both valid instruments: neither is correlated with ui4 – ui3. Proceeding in this manner, we can see that at T, the valid instrument set is (yi1, yi2,..., yiT-2). The matrix of instruments is W = [W' i,...,W' N]'. The moment conditions are given by E[(Δyit – γΔyit-1)yit-j] = 0 j= 2,...,t–1 t =3,..., T (9) or, in vector form, E (W' i Δ ui) = 0 (10) Where Δu' i = (ui3 – ui2,...,uiT – uiT-1) (11) Premultiplying (6) (here written in vector form) by W' gives W'Δy = W' (Δy-1)γ + W' Δu (12) Performing generalised least squares (GLS) on (12) gives the Arellano-Bond (1991) preliminary one-step consistent estimator:

γˆ1 = [(Δy-1)' W (W'(IN⊗G)W)-1 W' (Δy-1)]-1

[(Δy-1)' W (W'(IN ⊗ G)W)-1 W' (Δy)] (13) N where W' (IN ⊗ G)W = Σ W' i GWi and G is a (T-2) square matrix

i=1 with twos in the main diagonal, minus ones in the first subdiagonals and zeros otherwise [Baltagi 1995].

To check for the consistency of the above estimator following tests need to be performed. The estimator is consistent if there

Economic and Political Weekly June 30, 2007

is no second-order serial correlation in the error term of the firstdifferenced equation: it requires. E[ΔuitΔuit-2] = 0. A test for the validity of the instruments (and the moment restrictions) is a test of second-order serial correlation in these residuals. The most common test of the instruments is Sargan’s (1958) test of over-identifying restriction. It is asymptotically distributed as Chi-Square and tests the null hypothesis of the validity of overidentifying equations. The Wald test which checks for joint significance of all, or a subset of parameters is also performed. The null hypothesis refers to “insignificance”, implying that low p values suggest joint significance.

A common assumption in time series is that data are stationary. A stationary series has the property that mean, variance and autocorrelation do not change overtime. If the time series is not stationary then it can be transformed into a stationary series by first differencing. If the variables are stationary, the limiting distributions of most estimators will be approximately normal when T → ∞. If the data are non-stationary or contain unit roots, standard estimators have non-standard distributions as T → ∞. Unit root testing on both levels and first differences is done through stationarity tests designed specifically for panel data such as Levinlin and Ipshin. Levinlin et al (2002) estimates the panel unit root test developed by Levin, Lin and Chu. The test may be viewed as a pooled Dickey-Fuller test or an Augmented Dickey-Fuller (ADF) test when lags are included, with the null hypothesis of non-stationarity against the alternate that all series are stationary. Ipshin estimates the t-test for unit roots in heterogeneous panels developed by Im, Pesaran and Shin (2003). The Ipshin test assumes that all series are non-stationary under the null hypothesis against the alternative that at least one series is stationary. The results are reported in Section III.

III Supply Response Results

The regression results4 are reported for rice, wheat, sugar cane and bajra (Tables 4a to 4d). Levinlin and Ipshin tests indicate that few of the variables were found to be non-stationary in levels but none of them were found to be non-stationary after first differencing. Since Arellano-Bond dynamic panel data performs regression on differenced variables, time series properties are well satisfied (as on differencing all the variables were stationary). Arellano-Bond Dynamic Panel data regression assumes that there is no second order autocorrelation in the errors. Therefore a test for this is needed. Sargan’s test for overidentifying restrictions is used to check for mis-specification of the model. Failure to reject null hypothesis in both cases gives support to the model.


Table 4a presents the regression results for Panipat and Yamunanagar regions for the rice crop. These are the ones where water table is declining and percentage of rice area to gross cropped area (GCA) is also very high. The signs of the coefficients are as expected. Most of the coefficients are statistically significant. Depth (pre-monsoon) has a negative influence on acreage under rice while price affects acreage positively. Even lagged depth has a negative sign, though the coefficient is insignificant.

The null of Sargan test is not rejected which suggests that the model is well-specified. There is no autocorrelation (both first order and second order) present. Wald Chi square test of overall significance shows variables are jointly significant. However, the overall significance cannot be established. This may be because of very less number of observations (32) present, as only two regions are considered. The overall significance can be established if minimum support prices are used in the model instead of farm harvest prices.

The results for the water depleting zone and for the entire state (both zones) yield expected signs but were insignificant. In both these regressions (results not reported) depth and price have expected signs but are insignificant. This is the advantage of conducting a regional study, as concentration can be laid specifically on those regions which are growing rice in large amounts and at the same time also facing the problem of receding water

Table 4a: Results of Arellano-Bond Dynamic One-Step5 Estimation for Rice for Yamunanagar and Panipat

Using Farm Using Minimum Harvest Prices Support Prices

Constant 0.01 (0.00) 0.006 (0.00) Agca(t-1) -0.3 (0.01) -0.3 (0.01) Agca(t-2) -0.04 (0.06) -0.1 (0.04) Depth(t) -0.01 (0.00) -0.004 (0.00) Depth(t-1) -0.0002 (0.00) -0.001 (0.00) Price(t-1) 0.006 (0.00) 0 .0002 (0.00) Sargan’s test of over-identifying restrictions 1 1 Arellano-Bond test of first order autocorrelation 0.16 0.16 Arellano-Bond test of second

order autocorrelation 0.18 0.19 Wald Chi2 test 0.55 0.0006 No. of observations 32 32 No of groups 2 2

Notes: Agca denotes area under a particular crop as a percentage of gross cropped area. For rice the relative price with respect to (wrt) bajra is used. Depth of water table is for pre-monsoon period (June). Numbers reported for the tests are p-values.

Table 4b: Results of Arellano-Bond Dynamic One-Step Estimation for Wheat

Using Farm Using Minimum Harvest Prices Support Prices Depleting All Depleting All Regions Regions Regions Regions

Constant 0.001 0.002 -0.003 0.0001
(0.00) (0.00) (0.00) (0.00)
Agca(t-1) -0.07 -0.11 -0.05 -0.07
(0.06) (0.06) (0.05) (0.05)
Agca(t-2) 0.1 0.15 0.2 0.17
(0.11) (0.8) (0.1) (0.1)
Depth(t) -0.0001 -0.0007 0.003 -0.0003
(0.00) (0.00) (0.00) (0.00)
Depth(t-1) -0.004 -0.005 -0.003 -0.004
(0.00) (0.00) (0.00) (0.00)
Price 0.02 0.02 0.0002 0.00007
(0.01) (0.00) (0.00) (0.00)
Sargan’s test of overidentifying
restrictions 1 1 1 1
Arellano-Bond test of first order
autocorrelation 0.04 0.01 0.04 0.008
Arellano-Bond test of second
order autocorrelation 0.9 0.31 0.6 0.1
Wald Chi2 0.61 0 0.01 0
No of observations 80 160 80 160
No of groups 5 10 5 10

Notes: For wheat relative price is w r t to barley. Depth is for post-monsoon period (October).

table. The insignificant impact of pre-monsoon depth is apparent only when a group of regions is considered. When only the two rice growing regions are included, decreasing groundwater table has a clear impact on acreage under rice. The results for MSP reaffirm that the falling depth of water table would signal farmers to reduce the area under rice.


Wheat has been grown throughout Haryana. Hence results are reported for depleting zone and for all zones. The results obtained from regressions are mixed. Most of the variables have signs in accordance with expectations but very few are found to be statistically significant. Depth and lagged depth have negative sign but as they are insignificant it would be difficult to establish that depth has had a negative impact on its area under cultivation. The regressions using MSP for wheat also give similar results. Most of the coefficients are not significant. The coefficient of depth which was negative changes sign (though is still insignificant).

Sugar Cane

Yamunanagar, Panipat and Faridabad are the regions from water-depleting zone (which is the area of interest in this study) which have comparatively higher share of area under sugar cane. Hence results are separately reported for these. The signs of the variables are in conformity with the expectations. Coefficient of depth variable is negative and significant. Second, order autocorrelation is present in none of the regressions, implying that the estimates are consistent. Sargan test suggests that the model is correctly specified. Wald Chi square statistics is also favourable, suggesting that all variables are jointly significant.


The estimation is also done for bajra. The coefficient of depth in this case is non-negative. This suggests that farmers would not reduce acreage of this particular crop as water level falls further. Hence, even if water level is going deep farmers would not be discouraged from cultivation. Both depth and lagged depth have a non-negative sign. Relative price shows a positive sign. The diagnostic testing of the model also tells that the model is well-specified. Table 4d reports the results.

The results reported in this section firmly assert that farmers recognise the fact that water tables in some regions are declining and therefore they should adjust their cropping pattern accordingly and reduce the cultivation of water-intensive crops. As the water level falls farmers reduce the acreage under rice and sugar cane, both water-intensive crops. However, the acreage of wheat and bajra (the less water-demanding crops) does not fall in these regions.

But this brings the important question to the forefront that who are the farmers making these shifts. One can speculate that it must be the poor farmer who has to reduce the acreage of waterloving crops. Large farmers can continue to cultivate paddy as they may own high-powered electric pumps and still be able to extract water from low levels. Thus it can be assumed that it may be the affluent farmers who have started using high-powered pumpsets for irrigation, while the not so lucky may have shifted to cultivation of crops where water requirement is not so high. This may heighten the existing social and economic disparities, as it is the rich farmer who utilises the benefit of MSP offered by the government. This distorts the very purpose of providing MSP as a financial support to poor and vulnerable farmers.

IV Conclusion and Policy Recommendations

This section focuses on the main conclusions, limitations and recommendations of the study. The water problem in Haryana is distinct as both waterlogging and water depletion are observed. The study has noted with serious concern the rapidly declining groundwater levels in various parts of Haryana. It is all the more disturbing to find that the cultivation of water-intensive crops is increasing in the regions where the water table is falling. One of the causes for this is favourable MSP offered to farmers. As seen in Section III, farmers do respond to price signals and that prices are important factor in explaining farmer’s acreage behaviour. Besides this energy prices should also be altered so that they reflect the true cost of resource and hence prevent overexploitation of resources. The role of government can be of utmost importance in correcting distortion of resource use by appropriate pricing of both inputs and outputs.

Table 4c: Results of Arellano-Bond Dynamic One Step Estimation for Sugar Cane Using Farm Harvest Prices

Yamunanagar, Panipat and Faridabad

Constant 0.0005

(0.00) Agca(t-1) 0.7 (0.09) Agca(t-2) -0.2 (0.1) Depth(t) -0.003 (0.00) Depth(t-1) 0.003 (0.00) Price -1.09e-06 (0.00) Sargan’s test of over-Identifying restrictions 1 Arellano-Bond test of first order autocorrelation 0.15 Arellano-Bond test of second order autocorrelation 0.12 Wald Chi 2 0 No of observations 48 No of groups 3

Notes: For sugar cane absolute prices are considered. Depth is average of pre-monsoon and post-monsoon.

Table 4d: Results of Arellano-Bond DynamicOne Step Estimation for Bajra

Bajra Depleting All Regions Regions

Constant -0.003 (0.00) -0.003 (0.00) Agca(t-1) -0.4 (0.09) -0.2 (0.08) Agca(t-2) 0.2 (0.1) 0.2 (0.09) Depth(t) 0.004 (0.00) 0.003 (0.00) Depth(t-1) 0.002 (0.00) 0.002 (0.00) Price 0.01 (0.00) 0.02 (0.01) Sargan’s test of overidentifying restrictions 1 1 Arellano-Bond test of first order autocorrelation 0.1 0.1 Arellano-Bond test of second order autocorrelation 0.4 0.4 Wald Chi2 0 0 No of observations 75 153 No of groups 5 10

Notes: For bajra relative price is w r t rice. Depth is for pre-monsoon (June).

Economic and Political Weekly June 30, 2007

The decline in water table is confined not only to those regions where cultivation of water- intensive crops is more predominant. As seen in the study, Faridabad, Gurgaon and Mahendargarh are the regions where water table is receding but paddy cropping is not much (though increasing). This implies that the cause of declining water levels in parts of Haryana is explained by factors other than an increase in water-intensive crops.

Using the tool of supply response function the study establishes that depth of water table in a region has significant influences on planting decisions of water-intensive crops. Declining water table depth discourages the farmers to cultivate waterintensive crops. Farmers are indeed reducing their acreage of paddy and sugar cane in response to declining depth of water level over the years, while for dryland crops like bajra no such response has been displayed by farmers. Thus one is able to find a self-correcting mechanism of distorted cropping pattern in Haryana.

Several difficulties were encountered while undertaking this study. Some of those are pointed out briefly. Firstly, the regression analysis was done for 18-year period. The complete data relating to the study was not available beyond 1998-99 so that the econometric exercise could not be conducted beyond the said period. The study has used a sophisticated estimation technique. The robustness of the estimator could have been checked by using other price expectation formulations. But because of the short nature of time series this could not be performed. Secondly, the administrative district was chosen as basic unit for research analysis. Though, a study at block level would have helped us in understanding the problem better because the water depth varies markedly amongst various blocks in a district too. But such a study would have been unfeasible, as collecting consistent and reliable information at block level would have been difficult. Lastly, the relationship between depth and rice acreage runs both ways. It would also be interesting to look at both of these econometrically in a recursive system, where depth of the water has been affected due to cultivation of crops like paddy and sugar cane and the acreage of these crops is reduced to the falling depth of the water table. Besides the cropping patterns, the effect of electricity pricing on depth can also be analysed.




[This article is a part of the MPhil dissertation submitted at Delhi School of Economics. The author expresses her thanks to the supervisor, J V Meenakshi for her continuous support.]

1 Two important methods of charging for power supply: (a) flat rate system,

i e, a water extracting device is charged a monthly rate per horsepower

(hp) of the pumping plant regardless of actual power use; (b) pro-rata

charges, i e, a water extracting device is charged per unit (kWh) of power

consumed on the basis of meter readings 2 Almost 65 per cent of the agricultural area of Haryana state is underlain

by saline groundwater [Gangwar and Panghal 1989]. 3 This information is also collected by the Central Groundwater Board

(CGWB) in Delhi; however their information is based on a smaller number

of test tubewells and a complete time series was also difficult to construct,

hence this data source was not used. 4 The Arellano-Bond (1991) one step dynamic panel estimator is estimated

in STATA v 8.0 econometric software. 5 The first-step estimator is used since it has been shown to have more reliable

results. The asymptotic estimator from two-step GMM estimator has been

found to have a downward bias.


Arellano, M and Stephen B (1991): ‘Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations’, Review of Economic Studies, Vol 58.

Baltagi, H (1995): Econometric Analysis of Panel Data, John Wiley and Sons.

Bathla, S (1996): ‘Sustainability of Groundwater Resource in Punjab’, MPhil dissertation, Delhi School of Economics, Delhi University, Delhi (unpublished).

– (1998): ‘Sustainability of Land and Water Resources in Haryana: Some Ecology-Economy Interactions’, Report submitted to Agricultural Economic Research Centre (AERC), University of Delhi, Delhi.

Behrman, J (1968), Supply Response in Underdeveloped Agriculture: A Case Study of Four Major Annual Crops in Thailand, 1937-63, Amsterdam, North Holland.

Chand, R (1996): Agricultural Diversification in Punjab: Potentials and Prospects, Institute of Economic Growth.

Chand R and T Haque (1997): ‘Sustainability of Rice-Wheat Cropping System in Indo-Gangetic Region’, Economic and Political Weekly, Vol 32, No 13.

Cummings, J and H Askari (1976): Agricultural Supply Response, A Survey of the Econometric Evidence, Praeger, New York. Dhawan, B D (1982): Development of Tubewell Irrigation in India, Agricole Publishing Academy, New Delhi.

  • (1993): Trends and New Tendencies in Indian Irrigation, Commonwealth Publishers, New Delhi
  • (1995): Groundwater Depletion, Land Degradation, and Irrigated Agriculture in India, Commonwealth Publishers, New Delhi.
  • Duxbury, J M (2001): ‘Long-Term Yield Trends in the Rice-Wheat Cropping System: Results from Experiments and Northwest India’, Journal of Crop Production, Vol 3.

    Gangwar A C and B S Panghal (1989): ‘Trends in Groundwater Development and Management in Haryana’, study presented at workshop on Efficiency and Equity in Groundwater Use and Management, Institute of Rural Management, Anand, Gujrat.

    Hsiao, C (2003): Analysis of Panel Data, Cambridge University Press, New York.

    Im, K S, M H Pesaran and Y Shin (2003): ‘Testing for Unit Roots in Heterogeneous Panels’, Journal of Econometrics, Vol 115.

    Kataki, P, P R Hobbs and B Adhikary (2001): ‘Trend, Constraints, and Productivity of the Rice-Wheat Cropping System of South Asia’, Journal of Crop Production, Vol 3, No 1.

    Levin, A, C F Lin and C Chu (2002): ‘Unit Root Test in Panel Data: Asymptotic and Finite Sample Properties’, Journal of Econometrics, Vol 108.

    Prihar, S S, S D Khepar, Raghbir Singh, S S Grewal and S K Sodhi (1990): Water Resources of Punjab: A Critical Concern for the Future of Its Agriculture, Punjab Agricultural University, Ludhiana.

    Sargan, J D (1958): ‘The Estimation of Economic Relationships using Instrumental Variables’, Econometrica, Vol 26.

    Statistical Abstract of Haryana (1981-82 to 1999-00): Economic and Statistical Organisation, Planning Department, Government of Haryana.

    Economic and Political Weekly

    available at

    Alter Media Bookshop

    Brahmaswom Madom Bldg, M G Road, Trissur 680 001 Kerala Email:

    To read the full text Login

    Get instant access

    New 3 Month Subscription
    to Digital Archives at

    ₹826for India

    $50for overseas users


    (-) Hide

    EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

    Back to Top