Protection Structure and Total Factor Productivity Growth in India’s Organised Food Industry

Abstract

An evaluation of the level of protection accorded to primary agricultural products (Categories I and II) and processed agricultural products (Categories III and IV)—based on the data at four-digit level of the harmonised system (HS)—finds that the average applied tariff rate fell from 41% in 2001 to 35.3% in 2017, though with large variations (Chap. 14 of this book).

Appendices

Annex A

1.1 DEA Approach to Estimate TFP

Productivity is defined as the ratio of output to input(s). Productivity can be measured by using the single (or partial) factor productivity (SFP) and multifactor or TFP. The SFP is defined as the ratio of output (or value-added) to the quantity of factor of production. Three broad measures are used as follows:

• 1. Labour productivity, or output per labour;

• 2. Capital productivity, or output per capital and

• 3. Capital intensity, or capital per labour.

These partial measures of productivity do not capture the overall changes in the productive capacity of a firm or industry, and these are affected by changes in the composition of inputs.

The TFP represents the overall productivity or efficiency in input use. That is its advantage over other measures. An increase in TFP is termed TC. Several methods are used to estimate the TC and often include economies of scale and TE gains.

In a two-factor production model, the rate of growth of output is determined by capital stock, labour input and a “residual”, which is interpreted broadly as TC or technological progress in growth economics. The “residual” measures the rate of disembodied, autonomous, exogenous and neutral TC. The technical progress could also be embodied, induced, endogenous or non-neutral to factors of production. It could be labour-saving or capital-saving in nature (Goldar 1986; Balakrishnan and Pushpangadan 1994).

“Technological progress” and “TFP” are used interchangeably, but TFP captures only one aspect of technological progress—its effect on the overall efficiency of input use. TFP is considered to include both technological progress (regress) and change in TE. Technological progress is taken to be the advancement in knowledge relating to the art of production, which may take the form of new goods, processes or new modes of organisation. TE is defined as the efficiency with which factors of production are combined to generate output. While technological progress is conceptualised in terms of shifts in the production function (Solow 1957), the latter measures the distance between the actual and the frontier or maximum attainable levels of output (Bettesse 1990).

To estimate TFP, the literature has identified various techniques (non-frontier and frontier), which are classified into nonparametric and parametric approaches. Non-frontier techniques include the growth accounting and econometric methods. The frontier approach includes the Malmquist index, stochastic and deterministic frontier function.

In the parametric method, an explicit functional form is specified, and the parameters are estimated econometrically, implying that the estimates are sensitive to the selection of a particular functional form and the problem of simultaneity bias may arise. The nonparametric method does not impose any functional form; therefore, no direct statistical test can be performed (Goldar 2004; Das and Ghosh 2006; Kathuria et al. 2013).

The frontier approach aims to find the best obtainable position given the inputs or prices (Mahadevan 2003). It enables the decomposition of TFP into technical progress and TEC. The non-frontier approach assumes TE on the part of firms.

There is no consensus in the literature on the best approach for computing TFP. Many studies use non-frontier approaches, especially the nonparametric techniques. A few recent studies use the frontier approach and the parametric technique (Balakrishna et al. 2000; Marjit and Kar 2009).

The main methods are the growth accounting approach, multilateral TFP index, DEA analysis, Translog/Tornqvist index and the Levinsohn–Petrin method.

We employ the DEA approach, a nonparametric method, under the frontier approach, because it helps to overcome the shortcomings of the growth accounting approach, and it identifies the components of productivity change by decomposing the TFP index (Charnes et al. 1978). The TFP change is estimated using the Malmquist productivity index. It is defined as the ratio of two output distance functions. In other words, it measures the TFP change between two data points by calculating the ratio of the distances of each data point relative to a common technology. The variables usually used are cost of capital, labour, raw material consumed, energy used and gross value of output. Malmquist TFP index and efficiency scores are obtained by using the DEAP software.

The distance function is useful to describe a multi-input, multi-output technology without specifying the behavioural objectives of the firm (Kannan 2011). We use the concept of the output distance function to define the Malmquist TFP index. The output distance function measures maximal proportional expansion of output vector given input vector. Following Färe et al. (1994), for each time period t = 1, 2,…. T, the production technology set S^t consists of all feasible input vector $x^{t} \in R_{ + }^{N}$ and output vector $y^{t} \in R_{ + }^{M}$ such that x can produce y. The technology set is represented as follows:

$$S^{t} = \left\{ {\left( {x^{t} ,y^{t} } \right)} \right.: x^{t} {\text{can}} {\text{produce }}\left. {y^{t} } \right\}$$

The following output distance function is defined at t.

$$D_{0} \left( {x^{t} ,y^{t} } \right) = {\min}\left\{ {{\uptheta }:\left( {x^{t} , y^{t} /{\uptheta }} \right)\varepsilon } \right.{ }\left. {S^{t} } \right\}$$

$D_{0} \left( {x^{t} ,y^{t} } \right)$ represent the distance of a firm using x input vector to produce y output vector in period t relative to the reference technology in period t. The distance function $D_{0} \left( {x^{t} ,y^{t} } \right)$ will take a value of less than or equal to one if $\left( {x^{t} ,y^{t} } \right)$ is an element of technology set S^t. Further, $D_{0} \left( {x^{t} ,y^{t} } \right)$ will take a value of 1 if $\left( {x^{t} ,y^{t} } \right)$ is located on the boundary of the technology, and it will take a value greater than 1 if located outside the feasible technology set. The distance function is measured by using DEA, as in linear programming, by assuming constant returns to scale (Coelli 1996; Coelli and Rao 2003).

Following Färe et al. (1994), the Malmquist TFP index between period t and t + 1 can be represented as the geometric mean of output-oriented indexes: one using technology in period t as a reference technology and another using technology frontier in period t + 1 as the reference. The Malmquist TFP index is written as follows:

$$\begin{aligned} M_{0} \left( {x^{t + 1} ,y^{t + 1} ,x^{t} ,y^{t} } \right) & = \left[ {M_{0}^{t} \left( {x^{t + 1} ,y^{t + 1} ,x^{t} ,y^{t} } \right) \times M_{0}^{t + 1} \left( {x^{t + 1} ,y^{t + 1} ,x^{t} ,y^{t} } \right)} \right] \\ & = \left[ {\frac{{D_{0}^{t} \left( {x^{t + 1} ,y^{t + 1} } \right)}}{{D_{0}^{t} \left( {x^{t} ,y^{t} } \right)}} \times \frac{{D_{0}^{t + 1} \left( {x^{t + 1} ,y^{t + 1} } \right)}}{{D_{0}^{t + 1} \left( {x^{t} ,y^{t} } \right)}}} \right] \\ \end{aligned}$$

The Malmquist productivity index, defined in terms of distance functions above, evaluates whether the observed input/output combination has improved relative to reference technology in period t and relative to reference technology in period t + 1. The TFPG is positive if the value of productivity index (M₀) is greater than 1; negative if less than 1; and stagnant between periods t and t + 1 if 1. Following Färe et al. (1994), the Malmquist productivity index can be written as follows:

$$\begin{aligned} & M_{0} \left( {x^{t + 1} ,y^{t + 1} ,x^{t} ,y^{t} } \right) = \frac{{D_{0}^{t + 1} \left( {x^{t + 1} ,y^{t + 1} } \right)}}{{D_{0}^{t} \left( {x^{t} ,y^{t} } \right)}} \\ & \quad \times \left[ {\left( {\frac{{D_{0}^{t} \left( {x^{t + 1} ,y^{t + 1} } \right)}}{{D_{0}^{t + 1} \left( {x^{t + 1} ,y^{t + 1} } \right)}}} \right) \times \left( {\frac{{D_{0}^{t} \left( {x^{t} ,y^{t} } \right)}}{{D_{0}^{t + 1} \left( {x^{t} ,y^{t} } \right)}}} \right)} \right]^{{{1}/{2}}} \\ \end{aligned}$$

The component outside the square bracket is the ratio of TE in period t to TE in period t + 1. This efficiency change component indicates how far the observed production is getting closer or farther from the frontier. The expression inside the bracket indicates the shift in the technology frontier (TC) between the period t and t + 1. It is measured as the geometric mean of shift in technology between two periods evaluated at input levels x^t and x^t+1. If the value of efficiency change component is greater than 1, the production unit is catching up to the frontier in period t + 1 as compared to the period t. The improvement in TC provides evidence of innovation between two periods, and the value of TC greater than 1 shows technical progress.

Annex B

2.1 Construction of Variables and Sources of Data

The organised manufacturing sector in India comprises factories registered under Sections 2m (i), 2 m(ii) and 85 of the Factories Act, 1948. Under Section 2m, “factory” means any premises including the precincts thereof:

2 m(i) wherein ten or more workers are working or were working on any day of preceding 12 months and in any part of which a manufacturing process is being carried on with the aid of power or is ordinarily so carried on;

2 m(ii) where in 20 or more workers are working or were working on any day of proceeding 12 months and in any part of which a manufacturing process is being carried on without the aid of power or is ordinarily so carried on and does not include a mine subject to the operations of the Indian Mines Act, 1923, or railways run school. Under Section 85 of the Factories Act, 1948, the state government is empowered to notify any factory not covered under the above two sections.

The study is based on ASI annual data on organised manufacturing and National Sample Survey (NSS) quinquennial data on unorganised manufacturing sectors. It is important to mention that till 1988–89 the classification of industries was based on the National Industrial Classification (NIC) 1970 and from 1989–90 onwards based on NIC-1987 and then replaced by NIC-1998.

The Economic and Political Weekly Research Foundation (EPWRF) database has prepared concordance between NIC-1998 and NIC-1987 from 1976–77 to 2003–04. During 2000, the industrial classification changed to NIC-2004 and from 2008–09 replaced by NIC-2008. Many new industries have emerged and reclassified even at two-digit level. For instance, for long, manufacture of tobacco was considered as part of the food and beverage (NIC codes 20 and 21), which was taken to be a separate industry under NIC 2004 (code 16). Further, as per NIC-2004, code 15 represents food and beverage together which under NIC-2008 is changed to code 10 for food and code 11 for beverage.

To make a comparable time series on food and beverage industry in each state, a concordance matrix between three-digit classes of NIC-1998, NIC-2004 and NIC-2008 is prepared. From 1976–77 to 1997–98, data is extracted at NIC three-digit level based on NIC-1998 and from 1998–99 to 2013–14 from the Ministry of Statistics and Programme Implementation, Central Statistical Office. The industries falling under NIC three-digit level codes—151, 152, 153, 154 and 155—are categorised under processing of food and beverage. The industrial code 151 refers to production, processing and preservation of meat, fish, fruit, vegetables, oils and fats; 152 is the manufacture of dairy products; 153 is the manufacture of grain mill products, starches and starch products and prepared animal feeds; 154 is the manufacture of bakery and other food products; and 155 is the manufacture of beverages, including liquor.

Industry-Specific Variables

The following variables at nominal prices are extracted from the ASI processed data base, taken from the EPWRF database and the CSO.

Labour

Data on labour is obtained by adding production workers and non-production workers. As per the definition provided by the ASI, production workers relate to all persons employed directly or through only agency whether for wages or not and engaged in any manufacturing process or in cleaning any part of the machinery or premises used for manufacturing process. Persons holding positions of supervisor or management or employed in the administrative office, storekee** section and welfare section, engaged in the purchase of raw material, etc., are included in non-production workers. Total number of persons engaged is taken to represent labour.

Gross Fixed Capital Stock

Studies done on productivity show that the measurement of capital stock is afflicted by many conceptual problems, which explain the considerable differences in the estimation methodologies employed. The NAS makes benchmark estimates available, and, therefore, national-level estimates are easy to construct, but state-level estimation is hard and disaggregated industry-level estimation is harder still (for a full discussion see, among others, Roychaudhury (1977), Goldar (1986), Sarma and Rao (1990), Singh and Ajit (1995), Kumar (2001), and Sharma and Upadhyay (2008)).

The perpetual inventory method is most widely used in empirical research. For constructing a time series on gross fixed capital stock at constant prices, we require a series on gross investment, an asset price deflator, depreciation rate and a benchmark capital stock, and we adopt the following three-step procedure.

Step 1.

The most important prerequisite is the figure of capital stock in the benchmark (initial) year, i.e. K₀. To obtain K₀, it is assumed that the value of finished equipment of a balanced age composition would be exactly half the value of equipment when it was new. Hence, twice the book value of fixed assets in the benchmark year at 2004–05 prices is taken as a rough estimate of replacement value of fixed capital, i.e. K₀ = 2 × B₀ (where B₀ is the book value of fixed capital net of the depreciation in the benchmark year).

Step 2.

After obtaining the estimate of K₀, gross real investment (I_t) is obtained by using the following relationship:

$$I_{t} = \frac{{B_{t} - B_{t - 1} + D_{t} }}{{P_{t} }}$$

where B_t = book value of fixed capital in the year t, D_t = value of depreciation of fixed assets in year t, and P_t = implicit deflator for GFCF for registered manufacturing given in the NAS.

Step 3.

Given the estimate of K₀, the following relationship has been used to construct a series of gross fixed capital stock at 2004–05 prices:

$$K_{t} = K_{t - 1} + I_{t} + dK_{t - 1}$$

where K_t = gross fixed capital at 2004–05 prices in the year t, I_t = gross real investment in the year t, and d = annual rate of discarding capital. Following studies done on the subject, the annual rate of discarding of capital is taken to be equal to 5 per cent.

Measurement of Gross Value Added

The GVA is arrived at by deducting the cost of total input from the value of total output. The figures of “total output” comprise the total ex-factory value of products and by-products manufactured as well as other receipts from non-industrial services rendered to others, work done for others on material supplied by them, value of electricity produced and sold, sale value of goods sold in the same conditions purchased, addition in stock of semi-finished goods and value of own construction.

However, “total inputs” comprise the total value of fuels, materials consumed and expenditures such as cost of contract and commission work done by others on materials supplied by the factory, cost of materials consumed for repair and maintenance work done by others to the factory's fixed assets, inward freight and transport charges, rate and taxes (excluding income tax), postage, telephone and telex expenses, insurance charges, banking charges, cost of printing and stationery and purchase value of goods sold in the same condition as purchased. Rent and interest paid are not included.

The GVA in organised manufacturing total is deflated by the wholesale price index (WPI) of all commodities and food manufacturing by the WPI of food. A GVA measure based on a single deflation procedure may produce a bias in the estimates if material prices do not move parallel to output prices (Balakrishnan and Pushpangadan 1994; Kathuria et al. 2010), but this study uses the single deflation method because of the lack of state and industry-level deflators and other measurement problems in the use of the double deflation method.

Measurement of Gross Output

To obtain gross output, the figures of depreciation are added to net output. The ASI defines “depreciation” as the consumption of fixed capital due to wear and tear and obsolescence during the accounting year. It is taken as provided by the factory owner or is estimated on the basis of cost of installation and working life of fixed asset.

We use gross output figures instead of net output figures because depreciation charges in Indian industries are known to be highly arbitrary, as these are fixed by the income tax authorities and seldom represent true or actual capital consumption. However, the figures of net output consist of total value of all the products and by-products produced by the firm. This variable is deflated by the WPI.

Measurement of Wages (Emoluments)

The ASI provides data on emoluments paid to workers and includes all remuneration in monetary terms and also payable more or less regularly in each pay period to workers as compensation for work done during the accounting year. It includes direct wages and salary (i.e. basic wages/salaries and payment of overtime and dearness, compensatory, house rent and other allowances); bonus; paid holiday, etc.

The wages are expressed in terms of gross value, i.e. before deductions. Since we have taken the number of persons engaged, wage is represented by total emoluments and employer’s provident fund contribution. The monetary figures of emoluments have been deflated at constant 2004–05 prices using the consumer price index (CPI) of industrial workers. We use the splicing method to make a common base for the index. Data on deflators and other variables is extracted from NAS, CSO, Report on Currency and Finance, RBI, Handbook of Statistics on Indian Economy, RBI. Each variable specified in the functional form above is extracted from the ASI data at two-digit level for the organised foods industry.