1

Incorporating Regional Variations in a Macroeconometric Model for India: A Production Frontier Approach

By Shashanka Bhide and K.P. Kalirajan1

I. Introduction

Regional dimensions of a national economy attract attention of policy makers, researchers and investors, especially when there are large inter-regional differences in the level of development, growth and resources. In the Indian context, inter-state variations in economic growth and development have been a subject of continuing research and policy debate. Regional balance in development has been one of the major criteria in the formulation of various national policies. Assessing economic performance of the states, however, has not been as systematic as at the national level. For instance, macroeconometric or CGE models estimated for India through the last 50 years do not deal with the regional performance of the economy2. Even the official plan models used for the allocation of investment across sectors do not have a regional dimension. Approaches to incorporating regional dimension in an economy-wide model range from direct estimation of regional economy-models to desegregation of national level aggregates to regions based on existing regional shares. Implicitly, there is either an assumption of efficient allocation of resources across the regions or rigidities in the movement of resources across regions. An underlying constraint in building regional details in an economy-wide model has also been availability of adequate data. In this paper we present an attempt to incorporate regional dimensions in a macroeconometric model for India based on the Production Frontier approach to modelling output. The approach also provides a plausible link between policy and improvements in productivity. The macroeconometric models often incorporate a supply side of the economy based on a production function that relates output, at the aggregate or sectoral level, to the inputs, assuming a particular level of technology. The output is assumed to be the “maximum” possible for the given inputs and a technology. Technological change or progress shifts the production function such that for the same level of inputs, higher levels of output are obtained. Technological change is generally exogenous, with some exceptions where it is related to openness in trade, research and development expenditures and foreign investment. Production functions are specified at the aggregate or national level and the disagregation is with respect to sectors. As the production function approach is used at the aggregate level it does not examine variation in the technical efficiency in input use for different producers. The production frontier (PFR) approach on the other hand, focuses on variation in production performance across different producers. This paper presents the use of ‘production frontier approach’ to modelling the supply side in a macroeconometric model. The approach is applied to the case of Indian economy using state- level data on output of one sector, viz., agriculture, for estimating technical efficiencies, which

1 National Council of Applied Economic Research, New Delhi and Australian National University, Canberra, respectively. The paper is based on the work under a project ”Accelerating Growth through Globalization of Indian Agriculture sponsored by Australian Centre for International Agricultural Research, Canberra. 2 A number of extensive reviews of the macroeconometric modeling work on India are available. Two such reviews can be found in Krishna et al (1989) and Marwah (1991).

2

relate state- level production functions to the PFR. All the other sectors of the economy are modeled at the aggregate or national level. With the human capital development and infrastructure development as the possible means of improving efficiency, the model also provides an explicit role for policy to improve the growth potential of the economy at the state level.

In the next section of this paper we present a discussion of the methodology.

Development of the output block of the macro model, including the PFR based estimates for agriculture and manufacturing is presented in Section III. Remaining parts of the macro model are discussed briefly in Section IV. Selected applications of the model are presented in Section V fo llowed by the concluding remarks in Section VI.

II. Production Frontier Approach for Regional Level Modelling IIa. The basic framework The concept of production frontier is best understood in the context of a cross-section of producers. With a number of producers producing a homogeneous product using the same technology and the same inputs, they are likely to end up with different levels of output3. This variation in productivity can arise for a variety of reasons including differences in the quality of inputs, managerial factors, environmental factors or differences in regulatory environment under which production takes place. The assumption that all the producers use the same technology and same inputs may not hold strictly in practice. Put alternative ly, while there is a ‘potential’ level of maximum output that can be achieved from a given technology with given levels of inputs, individual producers may be able to achieve only a fraction of this potential for a variety of reasons. In applied empirical studies, the ‘potential’ maximum is obtained as an envelope of all the realized output levels across selected production units. The standard production function approach takes the ‘average’ response across the producers as the output that can be realized from given levels of inputs and the technology. The variation in performance across producers is, thus, averaged out.

The variation in production performance among the producers can be important if it is substantial. In the context of regions, such differences may arise due to factors that can be influenced by policy. For example, differences in the levels of infrastructure may influence costs of operation which may not be included explicitly as inputs and hence would lead to variation in output for the same level of measured inputs. In the production function approach, it is possible to estimate region-specific production functions if sufficiently detailed input-output data are available. In the absence of such information, an alternative is to use regiona l level data on inputs and output for estimating a PFR and associated regional level production functions. The basic framework for estimating a PF is the following specification for a production function,

Ln Qi = aoi+a1i LnX1i+a2i LnX2i + ε i…………(1) 3 Kalirajan and Shand (1994) present a discussion of the basic PFR approach and a number of applications in agriculture and manufacturing. In a recent paper, these authors (Kalirajan and Shand, 1999) have provided a survey of PFR approach.

3

Where Qi = Output for the i-th producer, Xji = Level of jth input for the ith producer, aij = parameters of the production relationship relating j-th input to output for the i-th producer, ε = random error term. The coefficients aji are assumed to be random with,

aji = a j + vji ………………………………….(2)

Where vji is distributed with mean zero and a constant variance; a j is constant that reflects the average response of output for variations in the level of j-th input. Note that the random error vji associated with the intercept term can be combined with the error term ε in (1).

Substituting (2) into (1) w get,

LnQi = a o + a 1 LnX1i + a 2 LnX2 i + wi……..(3) Where, wi = (ε i + voi + v1i LnX1i+ v2iLnX2i) …………. (4) E (wi) = 0 ………………….…….. (5)

Var (wi) =σ 2 + ∑=

2

1j

σ j (LnX)2

ij…… ……….(6)

Cov (wi, wi′) = 0 for i ≠ i′…………………… (7) σ j = var (aj) ……………………(8) In matrix form, Y= XB +w ………………….…………… (9) Where E(w) = 0……………………………………….(10) E(ww′) = Ω ………………………... (11)

Where, Y is a vector of output levels for n producers, X is a matrix of k inputs (including a column of ones) for n producers, B is a vector of k coefficients of production relationship, w is a vector of composite error terms (equation 4) and Ω is a (nxn) non-singular positive definite matrix.

Ω = diag (x1′ A x1, x2′ A x2, …. xk′ A xk)……(12)

Where A = E (aij - a j) (aij - a j) ′ ……………(13) The vectors xj have (nx1) dimension.

The statistical model in (3 –8) or (9-11) can be interpreted as a linear model

with heteroskedastic error term. Kalirajan and Shand (1996) adopting a methodology based on Hildreth and Houck (1968) and Griffiths (1972) show that along with a j, estimates of vji (in the case of v0i it is actually v0i+ ε i ) can also be uncovered. Thus, we have estimates of aji, providing producer-specific production function,

4

Ln Yi = a0′i + a1′i LnX1i + a2′i Lnx2i …………….. (14) Where aj′i are the estimated production function coefficients.

A production frontier is defined as, LnY* =a *

o + a *1 LnX1 + a2

* LnX2 ………..(15)

Where Y* = output from the production frontier, *ja = coefficients of the production

frontier such that *ja = maxaij ∀ i = 1, 2, ..n producers4.

Efficiency of a producer with respect to the frontier is defined in alternative

forms: overall efficiency and input-specific efficiency (Kalirajan and Shand, 1996). Overall efficiency is defined as the ratio of actual output of producer i to the output level from the frontier function (15),

OEFFi = (Yi /Y*)………………………………(16) Note that due to the stochastic nature of the frontier there is no restriction that

(OEFFi <1) in all the cases. However, if estimated OEFFi is defined as (^

Y i/Y*)

where ^

Y i is obtained as the predicted value of output from the production function for producer i, then (1 > OEFFi > 0).

Technical efficiency with respect to a specific input xj can be defined as, EFFXij = (aji /aj*) for j=1,2 ------------------------- (17) In the case of intercept, termed ‘general efficiency’, it can be defined as, GEFFi = (a0i / a0*) ----------------------------------(18) The approach also provides for a natural decomposition of output growth into

components that can be attributed to (a) input growth (b) change in technical efficiency and (c) technical progress (Kalirajan and Shand, 1998; Kalirajan and Obowana, 1998). This application is an important extension of the basic approach when we have time-series data on output and inputs on a cross-section of producers.

The production function is expressed for the panel data as, LnYijt = a0ijt + a1ijt LnX1ijt + a21jt LnX2ijt + ε it ……… (19) With,

akijt = (−a kjt + vikjt) ……(20)

4 Note that for convenience in discussion, we are ignoring the need to distinguish the intercept term in the original production function and the term when the function is transformed into the double-log form.

5

Following the specification for the case of cross section data, equation (20) can be re-written in a fashion similar to equations (3-8). Note that there is now a production function corresponding to each producer ‘i’ for each period ‘t’; production frontier can be defined for each period such that, Ln Y*

t = aot* + a1t

* LnX1t + a2*

t LnX2t …………….. (21) Where ajt* = max ajit ∀ i = 1,2,……n and t = 1, 2,…… t ……..(22)

The implication is that a technology frontier once achieved does not slip backwards. The movement from ajt

* to a*j (t + ∆) is due to technical progress. Deviation

of aijt from a*jt is due to the level of technical efficiency of the ith producer with

respect to the use of jth input in the t in period. Technical efficiency also may vary from one period to another for a number of reasons. IIb. Technical Progress:

Shifts in production frontier over time represent technical progress. A representation of the technical progress would be, (y *

t /y *1−t )

for specified levels of inputs. In a case where input-specific coefficients of production do not change significantly, technical progress can also be captured by the shifts in the intercept term associated with the production frontier, i.e., (a0t

* / a0 t-1* ) IIc. Determinants of technical efficiency Estimated technical efficiencies at the regional level are a link between the PFR and the production function. Although at a theoretical level, production relationship is specified such that output and inputs are homogeneous across producing units, in practice available data incorporates considerable heterogeneity in output and inputs. For example, in agriculture, output measured as value of gross output is an aggregate of the output of a number of diverse products. When output of a product increases by one unit, the gross value may increase at a different rate than when there is a unit increase in another product. The heterogeneity in inputs across producing units may arise because of differences in quality that are not taken into account at the time of measurement. One instance is the labour units. Differences in the skill level, ability are not incorporated, often, in the overall labour force estimates. Finally, efficiencies may also vary due to other factors such as the infrastructure facilities (roads, power supply, input/output marketing network, extension support in the case of agr iculture). Thus, from an empirical viewpoint, technical efficiency is the link between production function at the producer level and the frontier production function and shifts in PFR are due to technical progress. With cross section data for regions, one can estimate regional production function and with technical progress, each region may adjust at different speeds to the shifting production frontier.

6

III. Modelling the Output Block for the Macroecometric Model

The traditional disagregation of overall output of the economy has been in terms of agriculture (or primary sector), manufacturing, power, construction and services in the Indian macroeconometric models. In this application, we have desegregated overall output into (a) agriculture (b) manufacturing (c) mining (d) construction (e) power (f) trade, hotels and restaurants (g) transport, storage and communication (h) public administration and defense and (i) other services. The broad scheme for estimating sectoral output is provided in Table 1. The rationale for disagregation is the availability of data and the linkages that we seek to exploit in adapting the PFR approach to modelling output.

IIIa. Agricultural Output IIIa.1 Estimating Crop yield function There are four major components of agricultural and allied sector’s output in the Indian context: (a) crops, (b) livestock output (c) fisheries and (d) forestry. The crop sector accounts for 70% of the value of the entire sector, livestock 22%, fisheries 4% and forestry the balance of 4%. In this study, we estimate output of the crop sector in some detail and link the output of the other sectors to the output of the crop sector. There is considerable diversification of the crop output. Food grains, comprising of cereals and pulses account for 40% of the value of crop output, oilseeds 12% fibers 5%, fruits and vegetables 22%, sugarcane 7% and the other crops 14%. We have attempted to provide estimates of the shares of foodgrain and non-foodgrain crops in the value of crop output at the state level and also the combined share of rice and wheat in foodgrain output. The break-up of crop output into (1) foodgrain and non-foodgrain and (2) rice and wheat and other foodgrain is of policy interest in terms of price policy for food security.

The crop output at the state level is modelled within the framework of PFR approach outlined in the previous section. The crop output is first specified as a product of crop area and crop yield per hectare of crop area: Qjt = GAjt * Yjt ------------- (23)

Where Q is value of crop output, GA is the gross cropped area, Y is the crop

yield per hectare of crop area and the subscripts j and t refer to j-th state and t-th year, respectively.

The production/ yield function for j-th state in t-th year is specified as,

Ln Yjt = a0jt ′ + a1jt Ln R jt + a2 jt Ln (IA/GA) jt + a3 jt Ln (F/GA) jt + a4 jt Ln (LAB/GA) jt + a5 jt Ln (TR/GA) jt + a6 jt Ln RWFG jt -------------- (24)

Where R is the rainfall during June-September, IA is the irrigated crop area, F is fertilizer consumption LAB is the labor force in agriculture, TR is the number of tractors at the beginning of the year and RWFG is the ratio of rice and wheat to total foodgrain output. Q and GA are as described earlier. The subscripts ‘j’ and ‘t’ refer to j-th state and t-th year.

7

Data on a variety of variables were obtained from a number of sources and relevant variables were derived for the major 15 states for the period 1970-71 to 1992-93, on annual basis.

The production function coefficients (akjt) were estimated using the program TERAN developed at the Australian National University. The estimation procedure provides for a Lagrange-multiplier test for the ‘random coefficients model’ vs. ’fixed coefficients model’ of the production function. In the present case, the test rejects the null hypothesis of fixed coefficients model. The extent of variation in different coefficients is shown in Table 2. The Table provides the mean, minimum and maximum of the estimated input –specific technical efficiencies at the state leve l. These estimates suggest that the variation with respect to specific inputs is not large but the variation in the intercept is substantial: both across states and over time. Given the relatively narrow range of variation in the coefficients of specific inputs, we proceed by taking into account variation of the intercept alone for further analysis. Technical efficiencies are estimated for the intercept as, GEFF jt = (a0 jt / a0t *) -------------- (25)

Note that efficiency is estimated with respect to the PFR for a specific year ‘t’. We are not, thus, considering shifts in PFR. Based on the earlier decision to ignore variations in input –specific coefficients, the PFR for year ‘t’ is specified as5, Ln Yjt = Ln (a0t *) + .2489 Ln R jt + .1215 Ln (IA/GA) jt + .2178 Ln (F/GA) jt + .1244 Ln (LAB/GA) jt + .0731 Ln (TR/GA) jt + .2276 Ln RWFG jt ---- (26) The state-year specific production function is obtained by estimating GEFFjt as in equation (25) and applying it in equation (26) to get, Ln Yjt = Ln (a0t *. GEFFjt) + .2489 Ln R jt + .1215 Ln (IA/GA) jt + .2178 Ln (F/GA) jt + .1244 Ln (LAB/GA) jt + .0731 Ln (TR/GA) jt + .2276 Ln RWFG jt ---- (26a) Variation in GEFF across states provides a basis for distinguishing output response to different exogenous changes in the model. III.a.2 Estimating General Efficiency Equations

Taking into account data availability and plausible explanations for variation in technical efficiency across states following factors are hypothesized to be determinants of technical efficiency in Indian agriculture:

1. Ratio of food grain to non-food grain output 2. Ratio of agricultural NSDP to total NSDP 3. Ratio of agricultural NSDP to population 4. Ratio of NSDP from agriculture to manufacturing 5. Ratio of NSDP from agriculture to unregistered manufacturing 6. Rural literacy rate 7. Average farm size of the land holdings

5 In this and the next sections, we note some of the key estimated equations for clarity in discussion. The statistical estimates of these equations are presented in the Annexure.

8

8. Ratio of real NSDP from transport, storage and communications to total NSDP from all the sectors of the economy

9. Ratio of real NSDP from transport, storage and communications to population in the state.

The general formation of the regression model for overall technical efficiency

is GEFFjt = f (x1jit, x2ji…t)………………………..(27)

Where x1, x2, … are the explanatory variables chosen from the list above. As the formulation uses panel data, the GLS estimates would provide consistent and efficient estimates of the regression model. As the efficiency estimates need to be bound between zero and unity, we use a transformed version, viz.,GEFFjt/(1-GEFFjt) as the dependent variable rather than the GEFF itself. The transformation also implies, variability of the response in efficiency to changes in independent variables is dependent on the initial level of efficiency.

From alternative estimates, the selected equation for efficiency is,

LnGEFFjt/(1-GEFFjt)= 139.98 - .5215 Ln (FGQ/NFGQ) jt + .5514 DUMAGjt - 0.7596 Ln FSZjt + 0.0072 Ln FSZjt*RURLITjt

+0.00004 RURLITjt* (TSC/POP)jt – 0687 T …….(28) Where, (FGQ/ NFGQ) is the ratio of output of foodgrain to non- foodgrain; DUMAG is a dummy variable with value of 1 if the ratio of NSDP from agriculture to overall NSDP (both in real terms) is 0.4 or greater and zero otherwise; FSZ is the average farm holding size; RURLIT is the rate of rural literacy; (TSC/POP) is the per capita real NSDP from transportation, storage and communication and T is a trend variable. Ln is the logarithmic operator.

The estimated GEFF equation is a crucial equation in the model. It not only leads to variation in output for the same level of inputs across states, it also reflects state-specific characteristics that lead to variation in the production response of the states to inputs. Finally, it also captures the impact of changes in ‘output composition’ on the value of output through the variable (FGQ/NFGQ). This variable also provides a link between price policy for agriculture and the composition of crop output. III.a.3 Input demand functions Among the seven variables on the RHS of the yield function in equation (26), three are the input variables, viz., irrigated area (IA), fertilizer (F), labour (LAB) and tractors (TR). Of the remaining two, rainfall is an exogenous variable and RWFG is endogenously determined as a function of a state’s own characteristics, (IA/GA) and the ratio of price of rice and wheat to foodgrain price. Among the inputs, LAB is assumed to be exogenous. It may be more realistic to assume a mechanism where labor force in agriculture is a residual, after accounting for the labor requirements in the other sectors. However, given the simplistic coverage of the other sectors in the present model, it is more appropriate to model labor force employed in agriculture as an exogenously determined variable. We discuss below briefly estimation of the input demand equations. The estimated equations are not presented here to economize on space.

9

In the case of irrigation, we estimate a national level equation in which irrigated area is modeled in a manner similar to investment expenditures: as a function of public investment in agriculture (crowding- in effect) and the ratio of revenue per hectare from crops to price of manufactured products (terms of trade effect). The state- level area under irrigation is estimated from each state’s share estimated from its trend behavior.

Fertilizer consumption per hectare is estimated as a function of ratio (IA/GA), ratio of fertilizer price to lagged crop price and mean level of rainfall, all at the state level. Although we estimate only one equation, it captures state- level details through the explanatory variables. Purchase of tractors is estimated as a function of (IA/GA), ratio of price of tractors to lagged crop price, average size of land holdings, lagged crop output and lagged dependent variable. Stock of tractors is updated using an estimated depreciation rate and initial level of tractor stock. Total crop area, GA, is estimated as a function of IA, rainfall and lagged GA. Gross cropped area, fertilizer consumption and tractor purchase equations are estimated using panel of 15 states and 22 years’ (1971-72 to 1992-93) data, in each case. III.a.3 Agricultural output at the national level:

The estimated technical efficiencies are used to derive production functions at the state level. The national level output is calculated as a sum of the regional (state) outputs including the output from the regions, which are not modeled for lack of adequate data. The aggregate output is expressed as, QAjt = (Yjt

* . GEFFit) * GAit ……. (29) QAt = Σj=1,15 QAjt + QAot…………………..(30) QAot = f (Σj=1,15QAit)………………………(31) Where, to use slightly different notation, QA is the agricultural crop output (gross value in 1980-81 prices), with the subscript ‘j’ indicating j-th state, ‘t’ referring to t-th year and ‘o’ referring to ‘other’ regions not specifically modeled.

The information required for estimating Y*

it and GAit is obtained from the state level equations on input demand as well as relevant equations at the overall economy level. The linkages between regional level production and overall macro factors are bi-directional in the sense that variables such as prices, determined at the macro level, influence input demand and hence output while regional output adding up to total output, influences prices at the macro level.

For the remaining agricultural sectors, output (real value added) is first

estimated at the state level using equations to link crop output (value added) to the output from the other allied sectors. A link equation is also estimated to get real value added for crops, from the gross value of output. III.2 The Manufacturing Sector’s Output Manufacturing output is divided into output from the ‘organized’ and ‘unorganized’ sectors. In the case of the former, data on capital stock (K), labour employed (LAB) and value added (GVAD) are available from the Annual Surveys of

10

Industries. Data on these three variables for the period from 1970-71 to 1993-94 were used to estimate the production relationship6.

The basic framework for estimating the production function is specified in the PFR approach. Production function in a year ‘t’ is, Ln (GVAD/LAB)t = bot + b1t Ln(K/LAB)t + b2tt………….(32) Where the coefficients bo, b1, b2 are seen as random deviations from their respective means arising from changes in technology, demand related factors or technical efficiency. The variation associated with ‘bo’ has two components. One relates to the ‘general efficiency’ of the producers and the second, to the impact of demand factors that influence utilization of existing capacity. Variation in the coefficient b1 reflects partly changes in the efficiency of use of capital per worker and partly changes in technology. The variation in b2 reflects uneven technical progress. Technical progress is generally interpreted as a long-term process. However, there can be year-to-year or short-term fluctuations in the rate at which technology improvements take place, and these may well be related to the overall business environment. The year-to-year fluctuations, associated with the trend variable are distinguished here from the variations occurring around a stationary mean. In this sense, technical progress and its short-term fluctuations are measured here independently rather than as a residual.

The estimated equation indicates no significant variation in the coefficient b1 over time. The estimated equation7 for the organized manufacturing sector output is, Ln (GVAD/LAB)t = 0.3680 * θ0t+0.4148Ln(K/LAB)t+0.4157*θ2t.t……..(33) Where θ0 and θ2 are the efficiencies associated with intercept and time trend. The two are highly correlated with each other and we estimate the two variables as, ∆Ln θ0t = -0.1937 + 0.7202 ∆Lnθ0t-1 + 0.5004 *∆Lnθt-2

+3.4321∆ (STQ/MFGQ)t-1 +1.4143 ∆(STQ/MFGQ)t-2 -1.9721 ECMt-1……………………(34) ECMt= Lnθ0t – 3.4300 (STQ/MFGQ)t – 0.0027 t……………(34a) Lnθ2t = 0.0012 + 0.1131 Ln θ0t ………………………(35) Where STQ is the level of stocks at the beginning of the year and t is a trend variable. MFGQ is the gross value (real) of output of organized manufacturing sector. The specification attributes demand factors the sole influence on ‘general efficiency’ and through it on variation in technical progress. Equation (34) and (34a) are estimated as a co-integrating vector of variables. The level of stocks at the beginning of the year, which appears in the equations (34) and (34a) is estimated as, 6 Data on value of capital stock in constant prices are updates of the series given in Balakrishna and Pushpangadan (1994). 7 Details are in the Annexure.

11

(STQ/MFGQ)t = 0.0088 – 0.0014t + 0.2239 INFLMt + 0.1726 Ln (MFGQt/MFGQt-1)…….(36)

In order to estimate the value added from the entire manufacturing sector, we estimate an equation linking value added from organized sector to value added from unorganized sector. II.2a. Input demand Gross fixed capital formation and labour demand equations were estimated for the organized manufacturing sector. A combination of ‘accelerator’ and cost of capital explanation of investment demand is used to estimate, Ln GFCFt = -2.2179 + 0.4756 Ln Qt-1 + 0.4314 Ln GFCF t-1 -.0055 (NRt-INFLMt-1) …………………….(37) where GFCF= real gross fixed capital formation in organized manufacturing, NR = minimum lending rate of the financial institutions, INFLM = rate of change in the price index for manufacturing. The equation was estimated in the framework of Auto Regressive Distribured Lag Models (ARDL). Labour days employed in organized manufacturing is estimated as,

Ln LABt = 13.4582 + 0.1351t+ 0.2535 Ln (MFGQ t-1/MFGQt-2)

-0.0449 Ln NWt – 0.8740 Ln NWt-1- 0.2539 Ln NWt-2 -0.2717 Ln INFLCt + 0.2697 Ln INFLCt-1 + 0.1911 Ln INFLCt-2 + 0.3394 Ln INFLCt-3……….(38) Where NW is the index of nominal wage rate, INFLC is the rate of inflation based on consumer price index and all other expressions are as explained previously. The equation was estimated in the ARDL framework.

IV. Completing the Overall Model With the description of the output determination of the various sectors provided in the previous section, we discuss below the remaining major components of the model which provide the macroeconomic structure for the analysis of output effects at the regional level. The main components discussed in this section are price formation, fiscal variables, balance of payments block, and the monetary block. IV.1 Price formation IV.1a. Agricultural Prices

Agricultural Production as described in the previous section, is influenced by output prices via the input demand and output-composition equations. The agricultural prices entering into output or input demand equations are, (1) price of the composite commodity ‘rice and wheat’, (2) price of foodgrains (inclusive of grains other than rice and wheat), (3) price of non-foodgrain crops and (4) price index of all crops. Among the input prices, price of fertilizer is related to an energy price index, PFPL, which is exogenous, and the price of tractors is related to PM through link equations. Agricultural output prices are influenced, directly or indirectly, by the Government policies on the procurement of foodgrains from the farmers, distribution

12

of (subsidized) foodgrains to the consumers and the barriers on agricultural trade. The estimated set of price equations for agriculture is, Ln PRWt = - 5.7576 + .0072 t + .4867 Ln PPt + .2841 Ln PRW t-1 - .3234 Ln (QRW/ GDPR)t - .5907 Ln (QRW/ GDPR)t-1 ----------- (39) Ln POFGt = 3.8924 + .0445 t + .7646 Ln PPt -.2053 Ln QOFGt - .7705 Ln QOFGt-1 ---- (40) PFGt ≡ k1. PRW t + (1-k1)* POFGt ------------------- (41) Ln PNFGt = 2.5942 + .0815 t + .3566 Ln eRt + .0631 Ln PNFGt-1 - .5697 Ln PNFGt-2 + 0.1118 Ln (M1/GDPR)t - .3720 Ln (QNFG/GDPR)t ------------- (42) PA ≡ w1.PRW+w2.POFG+w3.PNFG ---------- (43) Where PRW is the price index of rice and wheat, PP is the index of procurement price for rice and wheat, QRW is the index of output of rice and wheat, POFG is the price index of other food grains, QOFG is the output index of other food grains, PFG is the price index of all food grains, PNFG is the price index of non- food grains, eR is the nominal exchange rate of the rupee, M1 is the narrow money stock, QNFG is the index of output of crops other than food grains, GDPR is real GDP and t is a trend variable.

The specification of agricultural prices also incorporates the impact of variations in overall money supply as well as the impact of variations in nominal exchange rate of the rupee. IV.1b Wage rates and price index for manufacturing Nominal Wage rate Nominal wage rate in organized manufacturing is related to output expectations, inflation rate and lagged wage rate as, Ln NWt = -2.0163 + 0.2653 LnMFGQt-1 + 0.7986 Ln NWt-1 +0.7654 INFLCt………(44) Where all the variables are as explained previously. Price of Manufactured Products Manufactured product prices are sensitive to both cost and demand pressures. As demand functions for manufactured products is not directly estimated, the price of manufactured products is estimated as a reduced form equation, Ln PMt = 3.3441 + 0.4188 Ln PFPLt + .0002 * [ (TAR * eR)t-1 +INDT t-1] + .3461 Ln PMt-1 -.4622 Ln PMt-2 + .3102 Ln (M3/GDP)t……(45) Where PM is the wholesale price index of manufactured products, PFPL is the wholesale price index of fuel, power, light and lubricants, TAR is the average collection rate of customs for manufactured products, eR is the nominal exchange rate (Rupees per US dollar), INDT is the rate of domestic indirect taxes of Central

13

Government per unit of value of output of manufacturing sector, M3 is broad money and GDPR is real GDP at factor cost. IV.2 Balance of Payments

Merchandise trade is modeled separately for agricultural and manufacturing sectors. Agricultural exports are specified in the form of a supply function where the role of world demand conditions have a positive impact on exports by reducing the cost associated with acquiring orders, accessing wider market for a variety of products and so on. In the case of rice, the world GDP is not the key factor given the niche market Indian rice enjoys. But government policy frequently interferes with the exports of rice of ‘common’ variety. Hence, in the export equations for rice, ratio of domestic price to international price and foodgrain stock with the government are taken as explanatory variables. The manufactured product exports are specified in the framework of a supply function in the sense that higher domestic prices relative to international prices reflect either greater profitability vis-a-vis exports or reduced competitiveness due to domestic cost increase. Exchange rate and export subsidies can influence exports by improving revenues from export sale relative to domestic sale. The equation for manufactured exports captures these features of trade. Imports of manufactured products other than petroleum products are modeled as demand determined: real GDP and the ratio of tariff-adjusted international price to domestic price are the explanatory variables. Agricultural imports are classified into three categories: imports of wheat, imports of pulses and imports of other agricultural products. The wheat imports are related to food price stabilization and food security concerns. Imports of pulses have become a regular feature as domestic production has fallen short of consumption demand. A time trend and output of other food grains (other than rice and wheat) provide an explanation of the imports of pulses. Net imports of petroleum products are estimated as a function of manufacturing output and domestic production of crude. Imports of petroleum products are positively related to manufacturing output and negatively to domestic production of crude. As manufacturing output increases, demand for energy increases. With a rise in domestic crude output, net import of petroleum products (including crude) is expected to decrease, ceteris paribus. Trade deficit is calculated as the difference between the value of merchandise exports and imports. Current account deficit is defined as the trade deficit plus deficit on invisible account. The latter is specified exogenously. Exchange rate is an exogenous variable in this model. IV.3 The Fiscal Block Fiscal variables affect the production sectors through both general mechanisms and direct interventions in the form of taxes, subsidies and expenditures in specific sectors. The general mechanisms of the transmission of the effects of fiscal imbalance on overall inflation rate, interest rate and exchange rate influence sectoral prices and output prices. In the present model as interest rate and exchange

14

rate are exogenous, the impact of fiscal imbalance through general mechanism is via inflation rate. Fiscal imbalance when monetized affects inflation rate and sectoral prices. Although monetization has become less important means of financing budgetary deficit, we have used this mechanism rather than use interest rate to balance the supply and demand for loanable funds. Choice of monetized fiscal imbalance to reflect the indirect impact of fiscal policy on the overall economy has also meant the need to focus more on Central government fiscal operations which are eligible for monetization, rather than the fiscal balances of the states. The Central tax revenues are related to manufacturing output, price of manufactured products and the rate of indirect taxes. Although revenues from import tariffs are not modelled separately, their variations are captured through the variations in manufacturing output. The non-tax revenues are taken as exogenous variable. Among the expenditures, three major items of current expenditures of the Central government are modelled explicitly: salaries and consumption expenditure; interest payments and subsidies. We estimate salaries and consumption expenditures in real terms as a function of fiscal stance of the government expressed as the output of public sector. Nominal expenditures on salaries and expenditures are obtained as a product of real expenditure and overall wholesale price index. Interest payments are related to the level of debt at the beginning of the year. Subsidies are estimated separately for fertilizers, food and the rest. Fertilizer subsidies are related to the ratio of energy price and fertilizer price and a trend variable. Food subsidies are related to the level of distribution of food grain through the public distribution system and the prices of procurement (purchase), distribution (sale) and the market conditions. The other subsidies in the Central government budget are estimated as a function for the ratio of energy price to overall consumer price index. Total expenditures include government investment expenditure (Central) and all other expenditures both the components being specified exogenously in nominal terms. Fiscal balance and monetized deficit of the Center are estimated using standard definitions: GFISCt = GEXPt - GREVt GBUDGCt = GFSCt – DBORt ′- EBORt ′ Where GFISC is the fiscal deficit, GEXP is the government expenditure and GREV is the revenue receipts; GBUDGCt is the monetized deficit of the Center, after netting out the borrowings (domestic: DBOR and external: EBOR) from the fiscal deficit. The borrowings are exogenously specified (hence the superscript′ ). IV.4 Monetary Block, the Consumer Price Index and the Inflation Rate The monetary block provides a link between fiscal balance (monetized), trade balance, prices and output. The specification abstracts from the influence of monetary policy on interest rate. IV.4a Monetary Aggregates Broad money M3, is estimated as a function of reserve money, RM and the inflation rate (consumer prices). Reserve money changes either due to monetization

15

of Central Government deficit or change in foreign exchange reserves. We relate narrow money M1 to broad money M3. IV.4b The consumer price index and inflation rate The overall price index at the wholesale level is estimated based on agricultural, manufacturing and energy (PFPL) price indices. Agricultural and manufacturing price indices are estimated endogenously whereas PFPL is an exogenous variable. The consumer price index (CPI) is related to the WPI and the ratio of narrow money to real GDP. Inflation rate is measured as the first difference of the logarithm of the consumer price index.

V. Some Applications: Regional Impact

As the macro model deve loped here has a focus on the agricultural sector, we consider two sets of policies that affect agriculture. One is the enhancement of agricultural productivity through measures that improve efficiency in production: improvements in rural literacy and rural infrastructure. The second relates to some macro level policies that impact on agriculture: exchange rate variations affect the trade volumes as well as prices. The impact on prices may not be uniform across sectors and the impact, therefore, may not be uniform across regions as well. The model provides an assessment of the impact at the regional level of the selected alternative scenarios of a macro nature or regional nature.

The impact of alternative scenarios can be measured by comparing the results

of the model “with the policy change” relative to the model results “without policy change”. The model can be solved for the purpose of simulation analysis either for the future periods or ‘within sample’ period. In the present study, we have restricted the analysis to ‘within sample’ simulations. The base run scenario (without policy change) is for the period 1975-76 to 1990-91. We have preferred the ‘within sample’ simulations, as the values of exogenous variables for the in-sample simulations are readily available. Secondly, a comparison of the impact for selected variables over a period of time does not indicate significant variation in the results for different time periods.

V.1 The initial conditions The simulations provide an assessment of the impact of alternative scenarios on the endogenous variables of the model. Due to non- linearities in the type of relationships, the impact of a change in some of the ‘exogenous’ variables is a function of the initial levels of endogenous/exogenous variables. The main variables whose initial levels are of importance in assessing the impact of alternative simulations are (1) those affecting general efficiency in agricultural production and (2) the proportion of irrigated area out of gross cropped area. For example, general efficiency in agricultural production is a function of rural literacy, infrastructure and a measure of diversification of crop output. The impact of rural literacy on efficiency, however, depends on infrastructure and the impact of infrastructure on efficiency is dependent on the level of rural literacy. Further, the impact of each of these variables in turn depends on the initial level of efficiency itself: lower the initial level of efficiency, greater is the impact. Hence the initial conditions of these variables are important in assessing the level of impact of alternative scenarios. With this in view,

16

we present the initial conditions of selected variables that are useful in examining the simulation results of the model.

State level variations in the levels of rural literacy are shown in Table 3 for the decadal Census years of 1971, 1981 and 1991. There is a large variation in literacy rates in all the three years, although there is improvement in all the states. Level of infrastructure for trade and commerce (per capita net state domestic product from transportation, communication and storage) also has large variation across the 15 states in all the three years (Table 4.). There is considerable change in the relative rankings, although some states (Bihar, Orissa) have been consistently poor performers and some (Tamilnadu, Maharashtra) have remained at the top of the scale. Estimated general efficiency also varies across the states. The lowest efficiency in TE 1992 (average for the three years ending in 1992) is for the state of Bihar at 58.10% and the highest is for Kerala at 93.67%. The extent of variation in the percentage of crop area that is irrigated in the years TE 1975, TE 1980 and TE 1992 is shown in Table 5. Again, the inter-state variation is large and there is considerable change for each state over time. The percentage of crop area that is irrigated is 62.31% in Punjab in TE 1992 (up from 56.67% in TE 1975) as compared to 16.42% in Maharashtra (up from 10.44% in TE 1974).

The large variations in initial conditions influence the impact of alternative

scenarios. Thus, taking 1991 as the base, the impact of a change in literacy rate on efficiency would be higher in Maharashtra as compared to Bihar, because the former has better infrastructure; the impact of improvement in infrastructure would be greater in Kerala than in Bihar, because the rate of literacy is greater in Kerala than in Bihar. Intuitively, a better ‘skilled’ work force can exploit the economic environment more effectively than a less ‘skilled’ labour force. Similarly, improving infrastructure in regions where labour force is more ‘skilled’ can yield greater benefits than in regions where labour is less ‘skilled’, of course keeping all other things equal.

A crucial point to be made is the initial level of general efficiency: in the states

of Kerala, Tamilnadu, Assam, the initial level of efficiency is relatively high and hence the net impact of improvement in rural literacy, infrastructure may not lead to higher crop output in comparison to the states of Maharashtra, Madhya Pradesh and Bihar where the initial level of efficiency is lower (Table 6). V.2 Impact of Rural Literacy and Physical Infrastructure on Agricultural Production We examine here the impact of rural literacy and infrastructure under three simulations: two simulations are carried out to assess the impact of improvements in rural literacy and TSC, separately; in the third simulation, both rural literacy and per capita TSC are increased by 10% over the base run levels. In the first simulation, SIMLIT, the rate of rural literacy is increased exogenously by 10% for all the states relative to the base-run values (with a cap of 100%). In the second simulation, the level of real NSDP from TSC per capita is increased by 10% in all the states. The results for simulation on rural literacy are summarized under SIMLIT and the results for improved infrastructure are under SIMTR. The combined impact of improvements in rural literacy and infrastructure are summarized under simulation SIMLT_TR.

The impact of increasing rural literacy and transportation infrastructure on

selected number of variables is summarized in Table 7 at the national level and in

17

Tables 7a-7b at the state level. The results are in terms of average percentage change per year for the selected variables relative to the base run scenario. V.2a. National level results In the case of an increase in rural literacy rate, across the states, by 10 percent, with a cap at 100%, crop output increases by 0.63 per cent per year. The increase in crop output results from the increase in general efficiency in crop production (by 0.59%). General efficiency raises crop yield per hectare that in turn leads to an increase in crop output. Increase in crop output, however, implies a reduction in agricultural prices (-0.31%). If there is no corresponding increase in the demand for crop output, higher level of production implies a drop in output price in order to raise consumption levels to absorb the higher output. As a second order effect, lower crop output price, relative to input prices leads to a reduction in the demand for inputs in crop production, if the input prices are not affected by the lower overall inflation rate. For instance, Table 7, indicates a reduction in fertilizer use (-0.28%) when there is an improvement in general efficiency due to the rise in rural literacy. Thus, although general efficiency improves, the level of use is lower in the case of some inputs. The lower agricultural prices, have a positive impact on manufacturing output as lower inflation rate, resulting from lower agricultural price, induces marginally higher fixed investment demand (less than 0.01%) and demand for labour (+0.1%) in the manufacturing sector. The labour employment in manufacturing improves as lower inflation rate reduces nominal wage rate. The general efficiency in manufacturing production, a measure that reflects the level of capacity utilization, increases marginally (+.02%) as a result of the drop in stock levels which in turn is caused by lower inflation rate. Thus, improvement in general efficiency in agriculture also leads to higher output in the manufacturing sector. The overall real GDP increases by 0.26 per cent in the case of a rise in rural literacy by 10%.

The impact of an improvement in infrastructure (increase in real NSDP from

transport, storage and communication, TSC per capita by 10%) for all the states, raises crop output in real terms by 0.31 per cent per year. The linkages by which the changes in efficiency, crop yield and output followed by the second order effects on crop prices and manufacturing output are effected are the same as described above in the case of increase in rural literacy.

A key factor influencing crop output is irrigated area. A change in irrigated area relative to total or gross crop area affects crop output in a number of ways. It influences crop yield directly, affects the use of other inputs such as fertilizers and purchase of tractors, influences total crop area under crops and it also has an impact on the cropping pattern. In the present model, cropping pattern is affected by irrigated area through its effect on the ratio of output of rice and wheat to total food grains and the ratio of non-food grain output to food grain output. The ratio of output of rice and wheat to total food grain output increases with a rise in irrigated relative to gross crop area. The ratio of food grain to non-food grain output increases with the rise in irrigated area relative to gross (total) crop area, following the impact of irrigated area on rice and wheat output. Thus, it is the food grain output other than rice and wheat that declines when irrigated area increases relative to total crop area. There may be an increase in both rice and wheat output and the output of non-food grains.

18

Irrigated area is affected by changes in crop output price relative to manufacturing price as well as by changes in crop yield relative to manufacturing price. In SIMLIT and SIMTR, irrigated area increases by .05% and .02% whereas gross cropped area increases by less than .01%. In other words, irrigated area as a proportion of gross cropped area increases under both the simulations although only marginally. Although crop prices decline when crop output increases, increase in the crop yield offsets the negative impact of lower crop price on irrigation. Similar impact is obtained on the purchase of tractors. Although crop output prices are lower, due to the higher irrigated area, tractor purchases are higher than in the base run. Increase in irrigation to some extent offsets the negative impact of lower crop prices on fertilizer consumption. In other words, while crop prices decline to induce demand for higher output, thus, thereby reducing the initial impact of enhanced efficiency on output, a number of interactions increase the input use and hence reduce the adverse impact of lower crop prices on input demand.

Lower prices are obtained, relative to the base run, in all the three crop groupings considered. However, the decline is the smallest in the case of non-food grain crops, followed by larger decline in the price of ‘rice and wheat’ group and largest decline coming in the price of ‘other food grains’. Changes in crop output also follow a corresponding pattern: output of non-food grain increases the greatest, followed by output of rice and wheat, and by ‘other food grains’. The variations in price response to output changes in different crop groups reflect differences in implicit price elasticity of demand. In the case of non-food grain crops, price elasticity is larger leading to smaller price response although output changes are greater. Increase in the output of non-food grain crops is caused by the sharper fall in the price of food grains relative to non food grains that implies a shift towards non-food grain production. Thus, the rise in efficiency initiated by the improvements in literacy and infrastructure is supplemented by crop diversification effect.

There is a decline in the amount of fertilizer consumption relative to the base run. This decline is due to the drop in crop price relative to the input price. In the case of fertilizer, as fertilizer price is held at levels determined by policy, drop in crop prices would adversely affect fertilizer consumption. In the case of tractors, although tractor prices also decline as they are linked to manufacturing price, the drop in crop price is greater. In the case of fertilizer, the adverse effect of lower crop prices is partially offset by increased irrigated area. In the case of tractor demand, increase in irrigated area compensates for the drop in the relative price of crops. We note that the drop in fertilizer (and the rise tractor demand) may be slightly overstated (understated) in the model as the direct impact of higher efficiency on fertilizer and tractor demand is not captured in the input demand equations.

Although agricultural exports increase and imports decline as a result of lower agricultural prices, the current account deficit increases as petroleum imports increase. Government revenues improve as manufacturing output increases and government expenditures decline. Both fertilizer and food subsidies decrease. Fertilizer subsidy decreases as fertilizer consumption declines and food subsidy decreases, as quantum of food distributed under the public distribution system is lower following the drop in market prices of rice and wheat. Food grain stock with the government increases as the level of food procurement rises and dis tribution is reduced. Although rice exports

19

increase, the impact of higher procurement and lower distribution on food grain stock is greater.

The impact of a combined increase in rural literacy and improved infrastructure are approximately additive. At the national level, crop output increases by 0.96% as compared to the increase of 0.63% under SIMLIT and 0.31% under SIMTR (Table 7). Similar nearly additive results are obtained in the case of all other variables of interest. The results point to the need for enhancing demand for agricultural products when policies shift the supply function upwards. If agricultural prices do not decline, the impact on input demand would not be adverse and the rise in crop output would be greater. However, the lower crop prices induce lower overall price leading to a rise in consumption demand for non-agricultural products. V.2b. State level results The model provides estimates of the impact on fertilizer consumption, purchase of tractors, efficiency and crop yield at the state level under alternative policy scenarios. Table 7a provides the estimates for the simulation in which rural literacy in all the states has increased by 10%. The impact of improved infrastructure at the state level is summarized in Table 7b. The effect of combined increase in rural literacy and infrastructure improvement at the state level is summarized in Table 7c. The state level variations are largely influenced by the changes in efficiency at the state level. Efficiency changes vary across the states as the response of efficiency to changes in literacy and transportation infrastructure depends on the initial level of efficiency itself and the levels of rural literacy and transportation infrastructure in each state. The response or elasticity is greater when initial level of efficiency is lower and the impact of literacy is greater when the infrastructure is superior. The impact is greater also when the initial level of literacy is higher.

Considering the impact of higher literacy on general efficiency first, the state in which the impact is the greatest is Maharashtra. MP, Haryana and Gujarat follow it. The impact is the lowest in Assam, Bihar and West Bengal. As noted in Table 4, level of per capita NSDP from transport, storage and communication (TSC) is the highest in the case of Tamilnadu followed by Maharashtra in TE1971 and TE1981. Therefore, the large response of efficiency to the increase in rural literacy in Maharashtra is partly explained by the relatively better infrastructure in the state. Gujarat and Haryana rank among the top 5 states in TE1981 and 1991. Bihar and West Bengal are among the bottom 3 states in terms of per capita NSDP from TSC in 1981 and 1991. The complementary effect of better infrastructure is, thus, greater in Maharashtra, Gujarat and Haryana and lower in Bihar and West Bengal. In the case of Assam, the initial level of general efficiency is among the highest (Table 5) out of all the 15 states considered. In contrast, MP has the third lowest level of general efficiency among the 15 states in TE1982. Maharashtra, Gujarat and Haryana have relatively lower levels of efficiency in the initial years of the simulation period. Thus, a combination of state level features determine the net impact of improvement in rural literacy rate on general efficiency in each state. The availability of adequate physical infrastructure, which in the present model is represented by TSC, is an important determinant of the impact of changes in rural literacy. The state level results for the impact of improvements in infrastructure (per capita TSC) are presented in Table 7b. The impact on general efficiency is the highest

20

in the case of Maharashtra, followed by UP and AP. It is the smallest in the case of Assam, Punjab and Rajasthan. As in the case of literacy, the impact is greater when initial level of literacy is higher, per capita TSC is greater and general efficiency level is lower. In the case of Maharashtra and UP, initial levels of TSC are relatively high and the initial levels of general efficiency are relatively low. In the case of AP the general efficiency has dropped over time significantly increasing its response to changes that affect efficiency. In the case of Rajasthan, although per capita TSC is relatively high levels of literacy are low in all the periods. Again, the determinants of net impact of improvement in infrastructure are initial levels of per capita TSC, rural literacy and general efficiency itself. The combined improvements in rural literacy and infrastructure result in the largest gains in crop output in the states of Maharashtra, Haryana and Gujarat (Table 7c). The impact is the least in the states of Assam, Bihar and West Bengal where the initial levels of parameters (literacy and infrastructure) are poor or efficiency is high. The results point to the positive impact of improvements in rural literacy and physical infrastructure on general efficiency in crop production which in turn result in higher crop output for the same level of inputs. The results across the states indicate that the states where initial levels of efficiency are relatively low are likely to respond more to policies influencing efficiency. The results also point to the interaction or complementarity between physical infrastructure development and human capital improvement. At the state level, therefore, policies aiming at improving literacy in rural areas and improving infrastructure for transportation are likely to result in significant growth of agricultural output. The variation across the states in the impact on fertilizer consumption and purchase of (demand for) tractors is indicated in Tables 7a and 7b for the two simulations, SIMLIT and SIMTR, respectively. In the case of fertilizer consumption, there is a decline in all the states when there is an improvement in efficiency. The drop in crop prices relative to the exogenously ‘fixed’ fertilizer price results in lower fertilizer consumption. The drop in fertilizer consumption is the steepest in MP, Maharashtra and Kerala. It is the least in the case of Haryana, Punjab and Tamilnadu. The estimated equations for fertilizer consumption suggest lower price elasticity in the states with higher irrigated to total crop area. The relatively larger irrigated area in Haryana, Punjab and Tamilnadu, thus, implies smaller impact of increase in the relative price of fertilizer. In the case of MP, Maharashtra and Kerala, irrigated area as a proportion of gross cropped area is among the lowest of all the 15 states. The purchase of tractors decreases in Assam, Bihar and West Bengal when rural literacy is increased by 10% whereas in all the other states, tractor purchases improve. Under SIMTR, tractor purchases decrease only in Assam. The drop in tractor demand is related to the fact that in the estimated tractor demand equation, tractor demand is positively related to the lagged value of crop output. Given the result that efficiency improvements are the least in the case of Assam, Bihar and West Bengal, crop output increase is also the lowest in these three states. Under SIMTR, the improvement in crop output is greater as the drop in crop price is less steep which implies smaller decrease in the use fertilizer per hectare. Thus, tractor demand is linked not only to the price of tractors relative to crop output price but also to variations in the level of crop output. The purchase of tractors increases by the highest

21

percentage under SIMLIT in Maharashtra, Haryana and MP. As efficiency improvements result in relatively large gains in output in the states of Maharashtra, Haryana and MP, purchase of tractors in these states is affected more than in the other states. The state level variations in the response to efficiency enhancing measures suggest that state-specific policies may be important to ensure maximum impact from such policies. The states where efficiency levels are presently low are likely to provide greater gains than the states where efficiency levels are relatively high. The states where both rural literacy and infrastructure are poor, the potential gains in efficiency are substantial. V.3 Impact of Devaluation of the Rupee The exchange rate variations affect agricultural sector through their impact on output prices and on the input prices. The transmission of the effect is, however, influenced by the restrictions on international trade and by rigidities in price adjustments. For example, fixed fertilizer prices at the subsidized levels are not affected by exchange rate variations unless the fertilizer prices are varied through policy measures. The trade restrictions in the case of food grains imply that price response to exchange rate variations in the case of non-food grains is likely to be greater than in the food grains. The estimated price equation for non-food grains includes exchange rate as an explanatory variable whereas in the equations for rice and wheat, and ‘other food grains’ exchange rate is not included as an explanatory variable. Nominal exchange rate also influences price of the manufactured products such that a depreciation of the rupee increases the price of manufactured products (PM). Thus, a depreciation (or appreciation) of the nominal exchange rate produces asymmetric impact across sectors depending upon the trade regime faced by each sector. In the present model, price of non-food grain crops and price of manufactured products increase relative to the price of food grain as a result of a depreciation of the rupee. The resulting changes in crop output reflect the altered pattern of price incentives to the producers. V.3a. National leve l results The aggregate or national level results of the simulation SIMER, where exchange rate of the rupee is depreciated by Rs 0.5 per US dollar relative to the base run are summarized for the national level in Table 7. First consider the price scenario resulting from the exchange rate depreciation. Price of non-food grains increases by 0.95% whereas the price of food grains increases by only 0.22%. Price of manufactured products increases by 0.50%. The changes in relative prices produce a corresponding output effect. Output of food grains decreases by 0.03% on account of the decline in the output of ‘other food grains’ by 0.13%. The non-food grain output increases by 1.02%. The output of rice and wheat increases marginally (+.02%) primarily due to the increase in irrigated area that raises the share of rice and wheat in food grain output. Although the rise in irrigated area relative to gross cropped area implies a reduction in the ratio of non food grain to food grain output, the positive impact of the rise in non food grain prices dominates the negative effect of irrigation expansion on non food grain production. Irrigated area increases by 0.07% as compared to a smaller increase of 0.01% in gross cropped area. Increase in irrigated area is a result of improved terms of trade as price of manufactured products increases by 0.50% as compared to the increase in

22

crop prices by 0.70%. The relatively higher increase in crop prices also induces higher demand for tractors (+1.86%) and fertilizer consumption (+0.60%). Aggregate crop yield increases by 0.38% and general efficiency increases by 0.14%. The rise in efficiency is the ‘crop diversification effect’ as the ratio of non-food grain to food grain output increases. Crop exports increase along with imports of agricultural commodities. Exports rise as price of exports in rupee terms increases relative to the domestic price. As domestic prices also increase, imports rise. Procurement of food grains by the government is projected to decrease, as the increase in procurement price (0.11%) is lower relative to the increase in the market price of rice and wheat (0.23%). Distribution through PDS increases by 0.09% and the food grain stocks with the government decrease by 0.81%. Thus, depreciation of the rupee is projected to result in higher agricultural prices and increased crop output. However, crop output mix is projected to change with larger proportion of non-food grain output than in the base run. With lower procurement and larger distribution through PDS, supplies of food grain with the government are likely to be smaller. Thus, in the longer run better targeting of PDS is important. The manufacturing sector shows a marginal change in output as a result of depreciation of the rupee. Although fixed investment increases modestly, there is a drop in employment as nominal wage rate increases. Real GDP from manufacturing of the organized sector increases marginally by less than 0.01%. The overall real GDP increases by 0.17%. The impact of rupee depreciation is projected to be favorable to the current account deficit, which is projected to decrease by about 7%. Thus, although domestic prices increase, the rise in export prices is relatively larger leading to increase in export earnings. The overall results suggest that agricultural production is likely to benefit from exchange rate depreciation more than the manufacturing sector especially when some of the input prices remain insulated from the effect of exchange rate changes. V.3b. State level results The state level results in Table 7d show that the crop diversification effect is the largest in Bihar (efficiency improves by 0.30%), followed by UP (0.28%) and West Bengal (0.26%). It is the least in the states of Assam (0.04%), Tamilnadu (0.06%), Gujarat (0.09%) and Kerala (0.09%). The pattern across the states is a result of the initially low levels of efficiency in Bihar, UP and West Bengal and the relatively high levels of general efficiency in Assam, Tamilnadu, Gujarat and Kerala. Thus, changes in cropping pattern are likely to result in greater impact on the crop output in those states where crop yield is relatively low for the same levels of input use. As level of efficiency is already at a high level in the states of Assam, Kerala, Tamilnadu and Gujarat, the change in cropping pattern does not increase efficiency as much as in the states with lower initial levels of general efficiency. The pattern of changes in fertilizer consumption and demand for tractors follows broadly the pattern noticed when all agricultural prices were increased under SIMPA. The per cent increase in fertilizer consumption is the highest in Maharashtra, MP and Kerala where irrigated area as a proportion to gross cropped area is the least;

23

the increase is the smallest in Punjab, Haryana and Tamilnadu where the coverage of crop area under irrigation is greater. The rise in purchase of tractors is the highest in UP, MP and AP and the least in Tamilnadu, Kerala and Assam. A combination of patterns of initial levels of irrigation (which influence demand for tractors positively) and the changes in general efficiency produced by exchange rate depreciation influences state level variation in demand for tractors. The efficiency gains are among the largest in UP and AP. And they are among the least in the case of Tamilnadu, Kerala and Assam. In the case of MP and AP, the ratio of irrigated to total crop area is relatively higher than the states with similar increases in efficiency. Bihar, MP and UP are projected to record the largest percentage gains in crop output as a result of rupee depreciation. The impact is the smallest in percentage terms for Tamilnadu, Assam and Punjab.

V. Concluding Remarks There are two sets of results from the study. One has a methodological

relevance and the other relates to policy implications of the results. In terms of methodology, application of the production frontier approach has been illustrated both where there is regional level data and when there is only national level time-series data. Application to regional level data is possible only when there are consistent estimates (across regions) on output and inputs. The source of inter-regional differences in production technology has been interpreted in terms of technical efficiency. While technical efficiency generally is with respect to individual or firm-level performance, aggregate performance reflects the influence of factors that may be beyond the control of an individual producer. An example of this is the influence of crop output composition on the aggregate output. The inter-regional differences in productivity as captured by the production function approach are not entirely due to differences in technical efficiency but also due to other factors. Specification of these other factors can provide links between policy and regional output performance.

The results of the study point to the importance of initial conditions for

states with respect to literacy, infrastructure and the ‘general efficiency’ to the impact of policies relating to these variables on crop output. The states that have a lower level of efficiency are likely to benefit more from measures that raise the level of efficiency. The impact of changes at the macro level influence output at the state level due to differences in the extent of coverage of irrigation, extent of crop-diversification as well as differences in general efficiency. From a policy perspective, an understanding of the impact of various policies at the regional level can be an important input in designing policies. The model presented here and its applications do have some important limitations. However, the paper illustrates an alternative approach to capturing the regional dimension in the overall assessments of the economy using macroeconometric models.

24

References Balakrishna, P. and K. Pushpangadan, “Total Factor Productivity Growth in Manufacturing Industry: A Fresh Look”, Economic and Political Weekly, Vol. 29, No.31, p. 2028-2035, July 30, 1994. Griffiths, W.E., “Estimation of Actual Response Coefficients in the Hildreth-Houck Random Coefficient Model”, Journal of American Statistical Association, 67, p. 633-635. Hildreth, C. and J.P. Houck, “Some Estimators for Linear Model with Random Coefficients”, Journal of American Statistical Association, 63, p. 584-85, 1968. Kalirajan, K.P. and R.T. Shand, Economics in Disequilibrium, An Approach from the Frontier, Macmillan India Limited, New Delhi, 1994. Kalirajan, K.P., M.B. Obwana and S. Zhao, “A Decomposition of Total Factor Productivity Growth: The case of Chinese Agriculture Before and After Reforms”, American Journal of Agricultural Economics, Vol. 78, No.2, p. 331-338, 1996. Kalirajan, K.P. and R.T. Shand, “Sources of Output Growth in Indian Agriculture”, Indian Journal of Agricultural Economics, Vol 52, No.4, p.693-706, 1997. Kalirajan, K.P. and R.T. Shand, “Frontier Production Functions and Technical Efficiency Measures", Journal of Economic Surveys, Vol. 13 (2), April 1999. Krishna, K.L., Krishnamurthy,K., Pnadit, V.N. and P.D. Sharma, “Macro-econometric Modelling in India: a selective review of research”, in Econometric Modelling and Forecasting in Asia, Development Papers No. 9, ESCAP, United Nations, New York, 1989. Marwah, K. “Macro econometric modelling of South East Asia: the case of India”, in Bodkin, R.G., Klein, L.R. and K. Marwah, editors, A History of Macro Econometric Model Building, Edwin Elgar, 1991.

25

Table 1 Scheme of Estimation of Output

Sector Approach 1. Agriculture Regional level,

production frontier approach 2. Manufacturing: organized sector

National level, production frontier approach

3. Manufacturing: unorganized sector

National level, Linked to organized sector’s output

4. Mining & quarrying National level, exogenous 5. Electricity, gas and water supply National leve l, exogenous 6. Transport, storage and Communication National, exogenous 7. Construction National level,

Linked to Current GDP and lagged construction output

8. Trade, hotels and restaurants National level, Linked to GDP from agriculture, manufacturing and exogenously specified sectors.

9. Public Administration and defense National level, exogenous 10. Other services National level,

Linked to current and lagged GDP and lagged output of own sector.

26

Table 2a. Variation in Efficiencies across States and over Time: Intercept and Irrigation

Intercept Irrigation State Mean Min Max Diff State Mean Min Max Diff Bihar 62.75 58.10 71.10 13.00 Kerala 95.91 95.67 96.28 0.60 MP 67.24 61.34 76.53 15.19 Assam 96.18 95.65 96.49 0.84 WB 71.92 67.95 74.04 6.09 TN 96.54 96.39 96.77 0.38 MT 72.52 69.76 77.40 7.64 Karnataka 96.81 96.33 97.05 0.71 UP 73.15 69.75 78.80 9.05 Punjab 97.04 96.88 97.38 0.50 Rajasthan 73.86 71.40 78.32 6.92 Haryana 97.12 96.97 97.32 0.36 AP 74.57 70.65 82.77 12.12 Gujarat 97.25 96.77 97.64 0.87 Orissa 74.64 71.00 76.37 5.37 UP 97.29 97.20 97.42 0.21 Gujarat 76.05 71.47 80.75 9.28 AP 97.32 97.26 97.42 0.16 Haryana 79.41 75.51 83.57 8.06 Orissa 97.40 97.14 97.74 0.60 Karnataka 81.13 77.50 85.32 7.82 Rajasthan 97.45 97.24 97.54 0.30 Punjab 82.95 80.14 86.26 6.13 WB 97.59 97.33 97.93 0.61 Assam 89.28 84.90 94.43 9.53 MT 97.77 97.04 98.78 1.73 TN 90.99 85.89 94.87 8.98 Bihar 98.08 97.94 98.17 0.23 Kerala 92.69 90.66 94.44 3.79 MP 98.29 97.74 98.59 0.85 Notes: (1) The state name abbreviations are: AP= Andhra Pradesh, MP= Madhya Pradesh, MT= Maharashtra, TN= Tamilnadu, UP= Uttar Pradesh and WB= West Bengal, (2) Min/max/mean/diff = minimum/ maximum/ mean and difference between minimum and maximum, respectively, of efficiency over the sample period 1970-71 to 1992=93. Table 2b. Variation in Efficiencies across States and over Time: Fertilizer and Tractors

Fertilizer Tractors State Mean Min Max Diff State Mean Min Max Diff Bihar 99.90 99.88 99.93 0.05 Bihar 99.60 99.50 99.74 0.24 MP 99.91 99.89 99.93 0.04 MP 99.65 99.54 99.78 0.25 MT 99.92 99.91 99.94 0.03 UP 99.72 99.67 99.80 0.13 WB 99.93 99.91 99.93 0.02 WB 99.72 99.67 99.76 0.09 UP 99.93 99.91 99.94 0.03 MT 99.72 99.66 99.79 0.13 AP 99.93 99.92 99.95 0.03 Rajasthan 99.73 99.68 99.79 0.11 Orissa 99.93 99.93 99.93 0.01 AP 99.74 99.69 99.83 0.14 Rajasthan 99.93 99.93 99.94 0.01 Gujarat 99.75 99.66 99.83 0.17 Gujarat 99.94 99.92 99.95 0.03 Orissa 99.76 99.71 99.78 0.08 Haryana 99.95 99.93 99.96 0.03 Haryana 99.81 99.75 99.85 0.10 Karnataka 99.95 99.94 99.96 0.01 Karnataka 99.82 99.78 99.87 0.10 Assam 99.96 99.95 99.96 0.00 Punjab 99.88 99.83 99.93 0.10 Punjab 99.96 99.95 99.97 0.02 Assam 99.91 99.86 99.96 0.10 Kerala 99.98 99.98 99.99 0.01 TN 99.93 99.88 99.97 0.09 TN 99.98 99.97 99.99 0.02 Kerala 99.94 99.93 99.96 0.03 Note: See notes for Table 2a.

27

Table 2c. Variation in Efficiencies across States and over Time: Labor, rainfall and output composition

Labor Rainfall RWFG State Mean Min Max Diff State Mean Min Max Diff State Mean Min Max Diff Bihar 95.92 94.11 98.14 4.03 Bihar 99.23 99.11 99.45 0.35 Bihar 92.86 91.62 95.10 3.48 Punjab 96.85 96.01 97.73 1.72 MP 99.34 99.18 99.58 0.39 MP 93.97 92.65 96.18 3.53 Haryana 97.20 96.55 97.52 0.97 WB 99.45 99.35 99.52 0.18 WB 94.84 93.87 95.67 1.80 AP 97.34 96.57 98.56 1.99 MT 99.46 99.39 99.57 0.18 MT 95.11 94.52 95.90 1.38 WB 97.36 96.68 98.55 1.87 UP 99.48 99.40 99.62 0.22 UP 95.11 94.40 96.47 2.07 Karnataka 97.37 96.97 98.07 1.10 Rajasthan 99.50 99.45 99.61 0.16 Rajasthan 95.25 94.64 96.28 1.65 Gujarat 97.44 96.69 98.56 1.87 Orissa 99.51 99.42 99.57 0.15 Orissa 95.36 94.50 96.06 1.55 UP 97.47 96.91 98.19 1.28 AP 99.51 99.43 99.69 0.26 AP 95.40 94.57 97.28 2.71 Orissa 97.62 97.04 98.62 1.58 Gujarat 99.54 99.43 99.64 0.21 Gujarat 95.51 94.26 96.33 2.08 MT 97.75 97.13 99.20 2.08 Haryana 99.61 99.53 99.71 0.18 Haryana 96.30 95.50 97.37 1.88 Assam 97.80 97.35 98.55 1.21 Karnataka 99.64 99.57 99.73 0.16 Karnataka 96.51 95.84 97.40 1.56 Rajasthan 98.00 97.27 98.70 1.44 Punjab 99.68 99.62 99.73 0.11 Punjab 97.02 96.39 97.66 1.27 MP 98.05 97.61 98.71 1.10 TN 99.80 99.71 99.89 0.18 Assam 98.19 97.32 99.03 1.71 Kerala 98.31 97.80 98.74 0.94 Assam 99.82 99.73 99.91 0.18 TN 98.35 97.41 99.37 1.96 TN 98.84 98.02 99.35 1.33 Kerala 99.89 99.84 99.95 0.11 Kerala 98.81 98.35 99.47 1.12 Note: See notes for Table 2a.

28

Table 3. Rates of Rural Literacy (%) across the Major Indian States State Year: 1971 State Year: 1981 State Year:1991 Rajasthan 16.44 Rajasthan 21.01 Rajasthan 30.37 Madhya Pradesh

20.08 Madhya Pradesh

24.62 Bihar 33.83

Bihar 21.013 Bihar 26.03 Andhra Pradesh

35.74

Uttar Pradesh 21.29 Andhra Pradesh

26.49 Madhya Pradesh

35.87

Andhra Pradesh

22.3 Uttar Pradesh 26.70 Uttar Pradesh 36.66

Haryana 25.92 Haryana 35.09 Orissa 45.46 Orissa 28.09 Karnataka 35.57 Karnataka 47.69 Karnataka 29.48 Orissa 35.70 Assam 49.32 West Bengal 30.63 West Bengal 37.90 Haryana 49.85 Assam 31.26 Assam 38.72 West Bengal 50.50 Punjab 32.02 Punjab 39.94 Punjab 52.77 Gujarat 33.31 Gujarat 41.46 Gujarat 53.09 Maharashtra 36.09 Maharashtra 43.47 Tamilnadu 54.59 Tamilnadu 37.04 Tamilnadu 43.54 Maharashtra 55.52 Kerala 68.54 Kerala 77.55 Kerala 88.92 Note: States are arranged in ascending order of % literacy in each year. Table 4. Level of Real Net State Domestic Product from Transportation, Storage and Communications Per Capita across the Major Indian States (Rs, 1980-81 prices) State Year: 1971 State Year: 1981 State Year:1991 Orissa 14.64 Orissa 19.10 Orissa 25.03 Bihar 19.38 Bihar 23.45 Bihar 33.30 West Bengal 22.22 West Bengal 25.97 West Bengal 42.98 Punjab 23.24 Punjab 33.27 Punjab 45.62 Assam 24.92 Assam 37.47 Assam 60.13 Madhya Pradesh

26.82 Madhya Pradesh

38.06 Madhya Pradesh

64.30

Kerala 38.48 Kerala 54.81 Kerala 79.47 Karnataka 38.81 Karnataka 55.53 Karnataka 85.01 Gujarat 41.77 Gujarat 63.02 Gujarat 90.98 Andhra Pradesh

43.96 Andhra Pradesh

63.04 Andhra Pradesh

106.57

Haryana 44.84 Haryana 70.81 Haryana 110.25 Rajasthan 52.12 Rajasthan 73.41 Rajasthan 147.96 Uttar Pradesh 52.52 Uttar Pradesh 75.57 Uttar Pradesh 176.75 Maharashtra 89.41 Maharashtra 116.58 Maharashtra 181.23 Tamilnadu 123.92 Tamilnadu 131.54 Tamilnadu 189.34 Note: States are arranged in ascending order of real per capita NSDP from transport, storage and communications in each year.

29

Table 5. Estimated General (Intercept) Efficiency in Agricultural Production across Major Indian States State TE1972 State TE1982 State TE1992 Maharashtra 63.58 Bihar 61.12 Bihar 58.10 Bihar 73.78 West Bengal 66.27 Madhya Pradesh 61.34 West Bengal 74.49 Madhya Pradesh 67.73 Maharashtra 69.77 Orissa 77.07 Uttar Pradesh 69.81 Orissa 71.00 Gujarat 77.71 Rajasthan 70.57 Gujarat 72.54 Madhya Pradesh 78.22 Orissa 74.77 Uttar Pradesh 72.74 Uttar Pradesh 80.59 Maharashtra 74.98 West Bengal 72.83 Andhra Pradesh 80.83 Andhra Pradesh 75.88 Andhra Pradesh 73.68 Karnataka 81.41 Harayana 76.96 Rajasthan 74.85 Rajasthan 81.91 Gujarat 79.22 Karnataka 78.45 Punjab 83.70 Karnataka 79.47 Harayana 81.09 Harayana 87.50 Punjab 80.55 Assam 84.90 Assam 89.75 Tamilnadu 86.61 Punjab 86.26 Kerala 95.42 Kerala 91.42 Tamilnadu 92.19 Tamilnadu 97.54 Assam 93.94 Kerala 93.67 Notes: 1. States are arranged in ascending order of general efficiency in each period. 2. TE refers to the three- year period ending in. Table 6. Irrigated Area as a % of Gross Cropped Area across Major Indian States State TE1972 State TE1982 State TE1992 Madhya Pradesh 8.73 Madhya Pradesh 11.92 Maharashtra 16.42 Maharashtra 10.44 Maharashtra 12.37 Kerala 18.73 Karnataka 14.78 Kerala 16.57 Madhya Pradesh 22.09 West Bengal 17.23 Karnataka 16.63 Assam 23.17 Gujarat 17.50 Assam 18.77 Karnataka 24.60 Assam 17.66 West Bengal 20.83 Orissa 28.24 Rajasthan 18.37 Orissa 21.11 Rajasthan 29.28 Orissa 18.53 Gujarat 23.07 Gujarat 30.90 Kerala 23.21 Rajasthan 24.16 West Bengal 37.34 Bihar 28.53 Bihar 33.46 Tamilnadu 39.51 Andhra Pradesh 34.00 Andhra Pradesh 36.60 Andhra Pradesh 41.39 Uttar Pradesh 37.37 Tamilnadu 41.06 Bihar 41.95 Tamilnadu 38.67 Uttar Pradesh 41.85 Punjab 62.31 Harayana 50.84 Punjab 56.09 Uttar Pradesh 65.42 Punjab 56.67 Harayana 59.84 Harayana 80.39 Notes: 1. States are arranged in ascending order of % irrigated area in each period. 2. TE refers to the three- year period ending in.

30

Table 7. Impact of Alternative Simulations on Selected Variables (Percentage change per year over the base run)

Impact under simulations Variables SIMLIT SIMTR SIMLIT_TR SIM_ER

I. Agriculture related Ia. Gross Value of Output Ia.1. Rice and wheat 0.5122 0.2608 0.7945 0.0243 Ia.2. Other foodgrain 0.5552 0.2473 0.8203 -0.1305 Ia.3. Total foodgrain 0.5061 0.2450 0.7704 -0.0324 Ia.4. Non-foodgrain 0.8099 0.3946 1.2361 1.0174 Ia.5. Total crops 0.6323 0.3072 0.9639 0.4038 Ib. GDP from Crops 0.5523 0.2674 0.8409 0.3592 Ic. Prices Ic.1. Rice and wheat -0.5057 -0.2392 -0.7614 0.2294 Ic.2. Other foodgrain -0.7278 -0.3082 -1.0542 0.1988 Ic.3. Total foodgrain -0.5613 -0.2564 -0.8346 0.2218 Ic.4. Non-foodgrain -0.1943 -0.0942 -0.2948 0.9528 Ic.5. Total crops -0.3089 -0.1447 -0.4633 0.7071 Id. Inputs Id.1. Fertilizer -0.2839 -0.1365 -0.4294 0.5958 Id.2. Tractors 0.5623 0.1911 0.7741 1.8639 Id.3. Irrigated area 0.0516 0.0238 0.0773 0.0075 Id.4. Gross crop area 0.0067 0.0028 0.0097 0.0103 Ie. Productivity Ie.1. Crop yield per ha. 0.6044 0.2892 0.9167 0.3843 Ie.2. General Efficiency 0.5900 0.2506 0.8591 0.1417 Note: Under SIMLIT, rate of rural literacy increased by 10% in all the states; under SIMTR, Per capita real NSDP from transportation, storage and communication increased by 10% In all the states; under SIMLIT_TR, the previous two simulations are combined and under SIMER, The exchange rate of Indian Rupee to US dollar depreciated by Rs 0.5.

31

Table 7. Continued.

Impact under simulations Variables SIMLIT SIMTR SIMLIT_TR SIM_ER

If. Trade If.1. Crop exports 1.2467 0.5686 1.9029 8.4777 If.2. Crop imports -0.7263 -0.3070 -1.0481 4.8155 II. Manufacturing: ASI IIa. Real Value Added 0.0600 0.0323 0.0948 0.0094 IIb. Real GFCF 0.0044 0.0021 0.0066 0.0659 IIc. Employment 0.0961 0.0502 0.1502 -0.0039 IId. General Efficiency 0.0184 0.0102 0.0294 0.0002 III. Macro Variables IIIa. Real GDP 0.2631 0.1269 0.4001 0.1668 IIIb. M3 -0.5359 -0.2425 -0.7955 0.404 IIIc. WPI -0.2272 -0.1050 -0.3398 0.4712 IIId. CPI -0.4097 -0.1882 -0.6114 0.3459 IIIe. CAD 0.5897 0.2677 0.8794 -6.9994 IIIf. Inflation rate (CPI) -0.0503 -0.0267 -0.0788 0.0485 IV. Other Variables IVa. Food subsidy -2.0102 -0.8245 -2.8766 -0.9836 IVb. Fertilizer subsidy -0.2310 -0.1059 -0.3441 0.6623 IVc. Monetized deficit -1.9567 -0.9261 -2.9424 1.8949 IVd. Fiscal deficit -0.3285 -0.1605 -0.4988 0.1966 IVe. Manufacturing Price -0.2317 -0.1067 -0.3464 0.4957 IVf. Terms of trade (PA/PM) -0.0772 -0.0380 -0.1168 0.2114

32

Table 7a. State Level Impact of Efficiency Improvement in Agriculture through Increase in Rural Literacy (by 10%): percentage change per year in selected variables over the base. State Fertilizer Tractor

(purchase) Tractor stock

General Efficiency

Crop yield Crop output

AP -0.2594 0.6358 0.2823 0.7519 0.7409 0.7659 Assam -0.3556 -0.1680 -0.0467 0.1118 0.0576 0.0669 Bihar -0.2882 -0.0742 -0.0197 0.1694 0.1301 0.1382 Gujarat -0.3118 0.7432 0.2141 0.8527 0.8271 0.8663 Haryana -0.1912 0.8676 0.3422 0.8916 0.8982 0.9566 Karnataka -0.3302 0.4935 0.1615 0.6873 0.6500 0.6715 Kerala -0.3861 0.0890 0.0110 0.3711 0.3142 0.3212 MP -0.4166 0.7611 0.3625 0.8955 0.8632 0.8997 Maharashtra -0.3876 1.2924 0.4683 1.3720 1.3469 1.3868 Orissa -0.3174 0.3647 0.1292 0.5589 0.5230 0.5378 Punjab -0.1966 0.6647 0.2117 0.7262 0.7220 0.7742 Rajasthan -0.3131 0.3831 0.1198 0.5482 0.5159 0.5443 Tamilnadu -0.2263 0.0785 0.0328 0.2833 0.2584 0.2735 UP -0.2616 0.3060 0.0904 0.4856 0.4616 0.4872 West Bengal -0.3418 -0.0725 -0.0192 0.1944 0.1459 0.1576 All -0.2839 0.5623 0.1928 0.5900 0.6044 0.6323 Table 7b. State Level Impact of Efficiency Improvement in Agriculture through Increase in Transportation Infrastructure (by 10%): percentage change per year in selected variables over the base. State Fertilizer Tractor

(purchase) Tractor stock

Efficiency Crop yield Crop output

AP -0.1249 0.2270 0.0850 0.3132 0.3042 0.3190 Assam -0.1702 -0.0957 -0.0237 0.0512 0.0255 0.0303 Bihar -0.1378 0.0512 0.0180 0.1655 0.1488 0.1543 Gujarat -0.1499 0.2094 0.0436 0.3069 0.2905 0.3189 Haryana -0.0917 0.2342 0.0730 0.3038 0.3000 0.3319 Karnataka -0.1565 0.0801 0.0230 0.2071 0.1857 0.1968 Kerala -0.1840 0.2191 0.0613 0.3552 0.3326 0.3355 MP -0.2010 0.0353 0.0158 0.1702 0.1439 0.1558 Maharashtra -0.1842 0.5444 0.1731 0.6219 0.6067 0.6369 Orissa -0.1509 0.0221 0.0021 0.1573 0.1360 0.1434 Punjab -0.0958 0.0327 0.0051 0.1334 0.1238 0.1381 Rajasthan -0.1511 0.0086 0.0006 0.1342 0.1147 0.1245 Tamilnadu -0.1092 0.1673 0.0635 0.2627 0.2538 0.2678 UP -0.1268 0.4152 0.1324 0.4766 0.4715 0.4956 West Bengal -0.1639 0.1106 0.0200 0.2333 0.2125 0.2231 All -0.1365 0.1911 0.0583 0.2506 0.2892 0.3072

33

Table 7c. State Level Impact of Efficiency Improvement in Agriculture through Increase in Rural Literacy (by 10%) and Transportation Infrastructure (by 10%): percentage change per year in selected variables over the base. State Fertilizer Tractor

(purchase) Tractor stock

Efficiency Crop yield Crop output

AP -0.3925 0.8902 0.3765 1.0935 1.0729 1.1142 Assam -0.5368 -0.2645 -0.0713 0.1671 0.0853 0.0998 Bihar -0.4351 -0.0103 0.0020 0.3507 0.2935 0.3077 Gujarat -0.4715 0.9639 0.2600 1.1742 1.1306 1.1991 Haryana -0.2890 1.1223 0.4209 1.2158 1.2183 1.3108 Karnataka -0.4969 0.5832 0.1867 0.9093 0.8487 0.8823 Kerala -0.5821 0.3295 0.0782 0.7534 0.6719 0.6820 MP -0.6306 0.8055 0.3811 1.0804 1.0196 1.0691 Maharashtra -0.5838 1.8709 0.6517 2.0307 1.9886 2.0605 Orissa -0.4782 0.3959 0.1331 0.7307 0.6717 0.6946 Punjab -0.2986 0.7027 0.2175 0.8675 0.8529 0.9201 Rajasthan -0.4741 0.3973 0.1213 0.6920 0.6388 0.6776 Tamilnadu -0.3427 0.2605 0.1012 0.5634 0.5289 0.5589 UP -0.3967 0.7690 0.2373 1.0085 0.9792 1.0314 West Bengal -0.5164 0.0568 0.0039 0.4502 0.3792 0.4025 All -0.4294 0.7741 0.2568 0.8591 0.9167 0.9639 Table 7d. State Level Impact of Depreciation of the Rupee by Rs 0.5 per USD: percentage change per year in selected variables over the base. State Fertilizer Tractor

(purchase) Tractor stock

Efficiency Crop yield Crop output

AP 0.5460 1.9221 0.8513 0.2226 0.4273 0.4481 Assam 0.7336 1.6727 0.4544 0.0418 0.2598 0.2727 Bihar 0.6267 1.9170 0.7459 0.2985 0.5133 0.5294 Gujarat 0.6353 1.7342 0.5345 0.0941 0.2967 0.3127 Haryana 0.4718 1.8700 0.8073 0.1290 0.3132 0.3390 Karnataka 0.7305 1.6670 0.5238 0.1207 0.3393 0.3537 Kerala 0.7966 1.6398 0.5725 0.0941 0.3349 0.3420 MP 0.7968 1.9323 0.8700 0.2062 0.4742 0.4993 Maharashtra 0.8236 1.7819 0.6317 0.1444 0.3938 0.4089 Orissa 0.7001 1.9003 0.6895 0.2242 0.4493 0.4656 Punjab 0.4228 1.8238 0.6800 0.0980 0.2620 0.2845 Rajasthan 0.6109 1.8094 0.6435 0.1072 0.3130 0.3303 Tamilnadu 0.4724 1.5590 0.5874 0.0554 0.2216 0.2363 UP 0.5382 1.9921 0.7707 0.2756 0.4749 0.4982 West Bengal 0.7115 1.9080 0.3999 0.2565 0.4667 0.4867 All 0.5958 1.8639 0.7041 0.1417 0.3843 0.4038

34

Annexure: Details of Selected Equations in the Model

This Annexure contains details of selected estimated equations that were referred in the paper. The details include method of estimation, estimated coefficients with their t-ratios, summary measures of R2 or mean squared error (MSE) and some diagnostic statistics. The Annexure does not provide information on all the equations of the model due to space limitations. The strategy adopted in the estimation differed for the purely time-series based equations and those which used ‘panel data’. For the equations based only on time-series data, we have used the Auto-regressive Distributed Lag Model approach of Microfit, which also provides a number of diagnostic statistics. In the selection of the equations, the tests for serial correlation, functional form, normality and heteroskedasticity were used to eliminate the ‘bad’ fits. However, in cases, where no other estimates were possible, equations were selected when the estimated coefficients were of the appropriate sign but some of the diagnostic tests suggested breakdown of the hypotheses on residuals. In the cases where we require only ‘link relationships’ such as the link between organized and unorganized sector output, we have used OLS estimates. For the panel data, the Generalized Least Squares approach is adopted when there is no lagged dependent variable among the explanatory variables. In the case of these exceptions, Instrumental Variables approach has been used for estimation. The notations used to indicate the variables are described in the text8. 1. The crop yield equation (mean response) Ln (Q/GA) = 1.3008 + 0.2489 Ln R + .1215 Ln (IA/GA) + .2178 Ln (F/GA) (2.87) (5.81) (1.97) (7.12) + .1244 Ln (LAB/GA) + .0731 Ln (TR/GA) + .2276 Ln RWFG (2.19) (3.09) (3.98) Method of estimation: GLS MSE = .0072 No. of observations: 345 χ2 (H0: fixed coefficients model) with 6 degrees of freedom: 29.85*** 2. Determinants of general efficiency in crop production LnGEFFjt / (1- GEFFjt ) = 139.9800 - .5215 Ln (FGQ/NFGQ) + .5514 DUMAGjt (16.45) (16.85) (15.13)

- .7596 Ln FSZjt + .0072 (Ln FSZ * RURLIT) jt (5.41) (2.51)

+ .00004 (RURLIT * TSC/ POP) jt - .0687 t (3.45) (16.25)

Method: GLS (Pooling of cross section and time-series data) R2 (Buse) = 0.6022 No. of observations: 345

8 Numbers within parentheses below the coefficient of the equation given here are the estimated t-ratios of the respective coefficients.

35

3. Production functions for Manufacturing: organized sector Ln (GVAD/LAB) = -1.2076 + .4148 Ln (K/L) + .0392 t (31.09) (8.90) (23.19) Method: GLS MSE = .0136 No. of observations: 23 χ2 (H0: fixed coefficients model) with 2 degrees of freedom: 4.15* 4. General efficiency in manufacturing: organized sector ∆ Ln θ0t = -.1937 + .7202 ∆ Ln θ0t-1 +.5004 ∆ Ln θ0t-2 + 3.4321 ∆ (STQ/QMFG)t-1 (6.04) (3.39) (3.46) (3.54) + 1.4143 ∆ (STQ/QMFG)t-2 – 1.9721 ECMt-1

(1.84) (6.65) ECMt = Ln θ0t – 3.4300 (STQ/QMFG) t-1 + .0027 t Method: ARDL (co- integrating vector) R2 = .8591 DW = 2.29 χ2 (SC)9 = 1.21 χ2 (FF)= 0.14 χ2 (N) = 1.77 χ2 (H) = 0.16 5. Stock levels of manufacturing output: organized sector (STQ/QMFG)t = .0088 - .0014 t + .2239 INFLMt +.1726 Ln (QMFGt / QMFGt-1) (2.09) (2.00) (3.08) (2.09) R2 = .4742 DW = 1.94 No. of observations: 23 χ2 (SC) = 0.01 χ2 (FF)= 0.43 χ2 (N) = 4.08 χ2 (H) = 0.22 6. Gross fixed capital formation (real) in manufacturing: organized sector Ln GFCFt = -2.2179 + .4756 Ln Qt-1 + .4314 Ln GFCF t-1 - .0055 (NRt – INFLMt-1) (1.51) (2.11) (1.77) (0.84) Method: ARDL R2 = .8462 DW = 2.21 No. of observations: 23 χ2 (SC) = 0.34 χ2 (FF)= 5.07** χ2 (N) = 1.53 χ2 (H) = 0.27 7. Labour demand in manufacturing: organized sector Ln LABt = 13.4582 + .1351 t + .2535 Ln (QMFGt-1/ QMFGt-2) - .0449 Ln NWt (42.38) (14.82) (1.37) (2.45)

- .8740 Ln NWt-1 -.2539 Ln NWt-2 - .2717 Ln INFLC t + .2697 Ln INFLC t-1 (3.10) (3.62) (2.36) (3.19)

+.1911 Ln INFLC t-2 + .3394 Ln INFLC t-3 (3.40) (1.85)

Method: ARDL R2 = .9943 DW = 2.40 No. of observations: 23 χ2 (SC) = 2.49 χ2 (FF)= 2.10 χ2 (N) = 1.30 χ2 (H) = 3.23*

9 The χ2 tests reported here relate to : SC= serial correlation, FF= functional form, N= normality and H= heteroscedasticity. Significance of the estimated statistics is indicated by *, ** and *** for 10%, 5% and 1% level of probability, respectively.

36

8. Nominal wage rate in manufacturing: organized sector Ln NWt = 1.4692 + .0437 t -.0424 D1t - .3660 Ln Qt-1 + .5808 Ln NWt-1 (3.45) (3.59) (1.02) (1.57) (4.76) + .6705 INFLCt (3.46) Method: ARDL R2 = .9953 DW = 2.47 No. of observations: 23 χ2 (SC) = 2.30 χ2 (FF)= 0.44 χ2 (N) = 4.30 χ2 (H) = 1.71 9. Price of manufactured products (wholesale price index) Ln PMt = 3.3441 + .4188 Ln PFPLt + .0002 [(TAR * eR)t-1 + INDTt-1] (7.27) (3.43) (3.74) + .3461 Ln PMt-1 - .4622 Ln PMt-1 + .3102 Ln(M3/GDPR)t

(1.79) (3.94) (4.82) Method: ARDL R2 = .9977 DW = 1.77 No. of observations: 23 χ2 (SC) = 0.05 χ2 (FF)= 2.38 χ2 (N) = 2.33 χ2 (H) = 0.02 10. Agricultural prices (wholesale price indices) Rice and wheat Ln PRWt = - 5.7576 + .0072 t + .4867 Ln PPt + .2841 Ln PRW t-1 (2.12) (.65) (2.56) (1.78) - .3234 Ln (QRW/ GDPR)t - .5907 Ln (QRW/ GDPR)t-1 (1.17) (2.07) Method: ARDL R2 = .9858 DW = 2.28 No. of observations: 24 χ2 (SC) = 0.93 χ2 (FF)= 1.44 χ2 (N) = 1.48 χ2 (H) = 2.33 Other food grains Ln POFGt = 3.8924 + .0445 t + .7646 Ln PPt -.2053 Ln QOFGt - .7705 Ln QOFGt-1 (2.00) (2.19) (2.58) (0.63) (2.51) Method: ARDL R2 = .9666 DW = 1.73 No. of observations: 24 χ2 (SC) = 0.30 χ2 (FF)= 3.00* χ2 (N) = 4.26 χ2 (H) = 1.90 Non- food grains Ln PNFGt = 2.5942 + .0815 t + .3566 Ln eRt + .0631 Ln PNFGt-1 - .5697 Ln PNFGt-2 (2.26) (3.40) (3.33) (0.36) (3.91) + 0.1118 Ln (M1/GDPR)t - .3720 Ln (QNFG/GDPR)t

(0.61) (1.44) Method: ARDL R2 = .9950 DW = 1.82 No. of observations: 24 χ2 (SC) = 0.16 χ2 (FF)= 1.84 χ2 (N) = 0.87 χ2 (H) = 1.14