What Really Affects Plant-level Productivity?

Abstract

This article investigates the plant-level characteristics that can affect plant-level productivity. Using Annual Survey of Industries (ASI) data from 1998–1999 to 2012–2013, we find that plant-level productivity is positively correlated with the number of products produced in the plant. It implies that the multi-product plants are relatively more productive than plants that produce only one product. We also find that plants are more productive if they enter the export market, and plant-level productivity is negatively associated with the ratio of production to non-production workers. This finding supports the fact that the plants with more skilled labour compared to unskilled labour are comparatively more productive. Additionally, we find that information and communication technology (ICT) has a significant and positive association with a plant’s productivity level.

Keywords

Productivity export multi-product fixed effect regression information and communication technology size India

JEL Classification: L11, O11, O25, O40, O53

Introduction

Over the past few years, India has experienced an increase in the growth rate of productivity in the manufacturing sector (Chatterjee & Jain, 2019). The pertinent question is to understand whether the internal factors like managerial ability and input choice have affected the productivity levels, and hence its changes over time. Most of the existing studies examining the productivity in the manufacturing sector have given little emphasis on the impact of plant-level characteristics on productivity in the context of India.

Majumdar (1997), using data of 1020 firms in India, found that larger firms, in terms of size, are less productive but earn more profits than smaller firms. More recently, Commander, Harrison, and Menezes-Filho (2011) found a positive association between information and communication technology (ICT) and productivity for 1000 firms in two developing countries like Brazil and India, using different econometric specification strategies.

Adding on to this strand of literature, this article seeks to focus on the plant-level characteristics that can affect plant-level productivity. Specifically, we consider whether plants are engaged in single or multi-product production, whether they are exporting or not, employing more of skilled workers (non-production) compared to unskilled workers (production), and if investing in computer equipment really improves plant-level productivity or not. Briefly, our objective in this article is to examine whether plant-level characteristics have any effect on productivity.

To analyse and comprehend these questions, we use 15 years of plant-level Annual Survey of Industries (ASI) data for the period from 1998–1999 to 2012–2013. To the best of our knowledge, this article is the first to explore the association between plant-level characteristics and productivity using the ASI panel data. We use fixed effect regression analysis to examine the relationship between plant-level productivity and its characteristics. Our findings are compelling. We found that exporting plants are more productive than non-exporting plants. However, the number of exporting plants is less, ranging from 7 per cent to 9 per cent in our data. The reason behind this small proportion of exporting plants could be the self-selection of firms which can bear the additional costs associated with the export market (Haidar, 2012; Wagner, 2007).

Next, we found that firms which diversify themselves by engaging in the production of multiple products are more likely to be much productive than firms that specialise in the production of only one product. Bernard, Redding, and Schott (2006) show that total factor productivity (TFP) is positively correlated with number of products produced and multi-product plants are more productive. In fact, they also find that the number of firms producing single product is more than the number of firms producing multiple products, which is consistent with our findings as well.

Studies have also found that information technology leads to productivity gains in the context of USA. (Jorgenson, 2001; Oliner & Sichel, 2000). In our data, plants have revealed their capital expenditure on computer software and equipment, which we use as a proxy for ICT. We find that ICT is significantly and positively associated with plant-level productivity.

Chuang and Lin (1999) in their study have shown that there is a significant impact of labour quality on the productivity of the manufacturing sector in Taiwan. Similarly, our result also shows that the plants with more non-production workers (skilled labour) compared to unskilled labour (production workers) are comparatively more productive. Moreover, we look at technology defined by capital intensity—the ratio of capital to total number of employees—and find that capital intensity is negatively associated with productivity, which can be due to capacity underutilisation.

The article has the following sections. The second section describes the data, the third and fourth sections give the descriptive analysis and regression results, respectively, and the fifth section concludes the article.

Data and Estimation Strategy

To understand the plant-level dynamics of productivity, we use the ASI panel data from 1998–1999 to 2012–2013 collected by the Central Statistical Office (CSO). The survey comprises of the following: (a) factories employing 10 or more workers with power and those employing 20 or more workers without power; (b) bidi and cigar manufacturing units registered under the Bidi & Cigar Workers Act; (c) all electricity manufacturing undertakings registered with the Central Electricity Authority (CEA); and (d) water supply, cold storage and repairing service units. Defence establishments, oil storage and distribution depots, restaurants, hotels, cafe and computer services and the technical training institutes, etc. are excluded from the purview of the survey. These factories are registered under Sections 2m(i) and 2m(ii) of the Factories Act, 1948.

Using the ASI data, we employ Hseih and Klenow (2009) methodology to calculate TFP. We calculate the productivity in terms of physical quantity (TFPQ).

A_{i s} = \frac{Y_{i s}}{L_{i s}^{1 - α_{s}} K_{i s}^{α_{s}}}

However, plants disclose their nominal output ( $P_{s i} Y_{s i})$ instead of real output (Y_is). To account for that, the following adjustment to productivity formula is required (Hseih & Klenow, 2009):

A_{i s} = \frac{P_{i s} Y_{i s}^{\frac{σ}{1 - σ}}}{K_{i s}^{α} L_{i s}^{1 - α}}

where A_is is the measure of productivity of a plant i in industry s, P_isY_is is the nominal output of a plant i in industry s, L_is is the wage payments of a plant i in industry s, K_is is the average capital employed by a plant i in industry s. Labour income share (1 − α_s) is the share of wage payments to gross value added in industry s at four-digit National Industrial Classification (NIC) industry level. Gross value added is the difference between the values of output and input.

Our measure of productivity is estimated at 4-digit NIC industry code. NIC is used to identify the economic activities of all the units. Over the years, there have been several changes in the NIC level. From 1998 to 1999 onwards, NIC 1998 was introduced and followed till 2003–2004. For the period from 2004–2005 to 2007–2008, the classification was made according to NIC 2004. Latest, NIC 2008 was used for the period from 2008–2009 to 2012–2013. The units are grouped into one of these NIC classifications based on their principal product irrespective of the number of products being produced by them. We have also evaluated concordance between NIC 1998 and NIC 2004 using NIC 2008.

We assume that production technology has constant returns to scale (Iyer, 2011). Due to this assumption, we exclude those industries whose labour income share is greater than 1 and less than 0. We limit our sample to those plants which have their log of productivity as non-negative. Our productivity measure is in logarithms, and we keep the plants with non-negative productivity.

We further use the information on plants’ geographical location, sectors (urban or rural), plant’s capital, depreciation, fixed assets, that is, land and building, furniture, fixtures, wages, number of factories, number of workers, employees, unpaid employees, quantity manufactured, quantity sold, ex-factory value of output, intermediate inputs, imported inputs and share of exports (in percentage) available in 8 different blocks of schedule. This information helps us to determine the plant-level characteristics.

Plants report their products and by-products which they produce during the year. This helps in segregation of single product and multi-product plants. Second, we estimate the size of the plant using the total number of employees employed by the factory manager measured in terms of logarithms. The total number of employees is defined as the sum of workers—directly employed or contractual workers, supervisors, managerial staff and other employees.

With rapid digitalisation across industries, we should expect a positive relationship between productivity and ICT. In order to validate this theory, we consider the data on investment in computer and its equipment including software and take it as proxy for ICT. We have taken the ratio of investment in computer and its equipment software to total capital investment by the plant. Capital intensity (in logarithm) is measured in terms of ratio of investment in capital to number of employees. The relationship between capital intensity and productivity is expected to be positive, as an increase in capital investment and decrease in the number of employees employed would lead to one-to-one increase in productivity.

Further, we calculate the ratio of production to non-production workers, It is used as a proxy for ratio of unskilled to skilled labour in our data. Production workers are engaged in routine operations and production process, whereas non-production workers are essentially managers who are responsible for supervision and decision-making. In our data, production workers are either employed directly by the plant or are hired on contract. On the other hand, supervisors and other employees comprise the non-production workers. Moreover, our data comprises of a share of products or by-products sold abroad (in percentage) by plants from 2008–2009 to 2012–2013. We construct a binary variable for exports which takes value 1 if the percentage share of products exported for a particular plant is greater than 0 and otherwise 0.

Now, we explore the relationship between plant-level characteristics and productivity using fixed effects regression estimation strategy:

log (A_{i s t}) = α_{0} + α_{1} {chr}_{i s t} + α_{2} {HHI}_{s t} + λ_{t} + δ_{i} + €_{i t}

(1)

where i denotes plant in industry s at year t, $A_{i s t}$ is the productivity of plant i in industry s at time t, ${chr}_{i s t}$ denotes the vector of characteristics of a plant i in industry s at year t. Specifically, it includes the size of the plant measured by the number of employees (in logs), the ratio of production to non-production workers, the capital intensity measured as the ratio of the capital to employees (in logs), the share of ICT to total capital, variable indicating whether the plant exports or not, and variable indicating whether it is a multi-product plant or not.

HHI = \sum_{i}^{n} {so}_{i}^{2}

where ${so}_{i}$ is share of output of plant i in industry s. We also include time fixed effects measured by $λ_{t}$ to control changes which might have affected all the plants simultaneously. Other than variables identified as plant-level characteristics, there could be few unobservable variables like management ethics and plant’s culture which might be correlated with productivity. To control for these time-invariant unobservable variables, we include plant fixed effects, $δ_{i}$ (Chi, 2005).

Descriptive Analysis and Systematic Productivity Differences

In the context of India, Chatterjee and Jain (2019) have found that TFP has grown 8.9 per cent using ASI data, which covers approximately 100,000¹ production plants over the period from 1983–1984 to 2012–2013². To start with the variables in interest, we try to find the systematic differences in productivity across plants. Melitz (2003) has shown that trade helps firms to become more productive and enter the export market. Trade persuades firms to reallocate its resources and increase aggregate productivity. Moreover, the least productive firms exit the market while the rest still go on producing to serve domestic firms. De Loecker (2011) stressed that abolishing trade barriers increases productivity. Consistent with the findings of these studies, we also find that exporting plants are more productive than non-exporting plants across time, that is, from 2008–2009 to 2012–2013 (see Table 1). However, the number of exporting plants is less than the number of non-exporting plants (see Figure 1). There are additional costs associated with exports, which are over and above the production costs. Plants which are able to cover these costs are able to enter the export market (Haidar, 2012; Wagner, 2007).

Table 1.

Productivity Differences Between Exporting and Non-exporting Plants

Year	Exporting	Non-exporting	Overall	Difference
2008–2009	12.039	11.28	11.346	0.758***
				[0.033]
2009–2010	11.933	11.183	11.241	0.750***
				[0.032]
2010–2011	12.921	11.339	11.339	1.582***
				[0.536]
2011–2012	12.127	11.339	11.41	0.788***
				[0.03]
2012–2013	12.375	11.59	11.651	0.784***
				[0.045]

Source: Authors’ own work using ASI data.

Note: Standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

Figure 1.

Source: Authors’ own work using ASI data.

In our data, plants have reported the information on a number of products they are producing in a year. It helps us to bifurcate the plants into two groups: group one comprises the plants which produce only one product and group two consists of plants producing more than one product, maximum up to 11 products. We find a significant productivity difference between the two groups of plants (see Table 2). It is expected that there are rewards for specialisation and single product plants should have higher productivity. On the other hand, there might be learning spill over from one product to another, giving multiple products plants producing an edge. Our finding suggests that the second effect is dominant. Bernard et al. (2006) also find that TFP is positively correlated with number of products they produce and multi-product plants are more productive. We also find that the number of firms ranging from 42 per cent to 51 per cent and producing single product is more than the number of firms ranging from 36 per cent to 45 per cent and producing multiple products, which is consistent with our data.

Table 2

. Productivity Differences Between Multi-product and Single Product Plants

Year	Multi-product	Single Product	Overall	Difference
1998–1999	11.611	10.8	10.997	0.811***
				[0.031]
1999–2000	11.612	10.811	10.992	0.801***
				[0.03]
2000–2001	11.501	10.755	10.906	0.747***
				0.027
2001–2002	11.529	10.738	10.944	0.791***
				[0.031]
2002–2003	11.588	10.837	11.006	0.751***
				[0.03]
2003–2004	11.411	10.678	10.827	0.733***
				[0.023]
2004–2005	11.602	10.787	10.983	0.815***
				[0.021]
2005–2006	11.61	10.82	10.996	0.790***
				[0.021]
2006–2007	11.756	10.962	11.109	0.794***
				[0.033]
2007–2008	11.95	11.23	11.365	0.720***
				[0.028]
2008–2009	11.959	11.239	11.346	0.720***
				[0.025]
2009–2010	11.841	11.171	11.241	0.670***
				[0.022]
2010–2011	11.978	11.261	11.339	0.717***
				[0.022]
2011–2012	12.069	11.315	11.41	0.754***
				[0.021]
2012–2013	12.276	11.534	11.651	0.742***
				[0.034]

Source: Authors’ own work using ASI data.

Note: Standard errors are in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.1.

Regression Analysis

In this section, we examine the relationship between plant-level characteristics and plant-level productivity using Equation 1. The results from this regression are presented in column (1) of Table 3. As discussed earlier, multi-product plants are more productive compared to single product plants. In the regression analysis, we find that multi-product product plants seem to have 10 per cent higher productivity compared to single product plants. This result is consistent with the fact that TFP is positively correlated with number of products produced by firms (Bernard et al., 2006). Further, we find that the size of the firm, measured by the number of employees, is significantly and positively correlated with plant-level productivity. The coefficient indicates that 1 per cent increase in size is associated with 38.9 per cent increase in plant-level productivity.

In the era of digitisation, we consider ICT as one of the factors that can improve productivity. Our data provide the amount of capital invested in the form of computer equipment. Our fixed effect regression analysis shows that an increase in share of investment in computer equipment in relation to total fixed capital seems to improve the productivity by 11.1 per cent.

Next, we look at technology defined by capital intensity—the ratio of capital to total number of employees. The coefficient in Table 3 indicates that the intensity is negatively correlated with the productivity. One per cent increase in capital intensity will decrease productivity by 33.9 per cent. One of the possible reasons for this could be underutilisation of capacity installed in the plants. Underutilisation implies that there is a gap between installed capital, which is measured, and actual capital used in production. Studies have pointed out that there was an investment boom in the 1990s. However, these studies have also shown that capacity utilisation rate is lower despite improvement in capacity utilisation. This is due to the shortage of demand as well as rigid labour laws and it has a negative impact on productivity (Deb, 2014; Dougherty, 2009; Dougherty, Herd, & Chalaux, 2010; Goldar & Kumari, 2003; Uchikawa, 2001).

The effect of labour quality, that is, skilled labour versus unskilled labour, on productivity is not much explored in the literature. Chuang and Lin (1999), in their study, have shown that there is a significant impact of labour quality on productivity of the manufacturing sector in Taiwan. They have defined labour quality in three ways: the ratio of skilled to unskilled workers, the wage of white-collar to blue-collar workers and share of white-collar workers who are employed. In our analysis, we observe the ratio of production workers to non-production workers is negatively associated with productivity (see Table 3). Moreover, skilled labour increases the firm’s efficiency by raising the firm-level productivity as illustrated by Chuang and Lin (1999). Employing a higher proportion of skilled labour (non-production worker) is a part of investment in human capital. Our finding is also consistent with the fact that skilled labour drives productivity.

Table 3

. Regression Results

Variables	[1]	[2]
Variables	Productivity (in logs)	Productivity (in logs)
Multi-product	0.100***[0.00]	0.076***[0.00]
Size	0.389***[0.00]	0.371***[0.00]
Share of computer capital to total fixed capital	1.111***[0.02]	0.777***[0.04]
Log of capital to employees	−0.339***[0.00]	−0.323***[0.00]
Ratio of directly employed workers to supervisors and others	−0.006***[0.00]	−0.004***[0.00]
Exports		0.027***[0.00]
Plant fixed effects	Y	Y
Year fixed effects	Y	Y
Observations	14,88,288	6,03,387
R-squared	0.903	0.954

Source: Authors’ own work using ASI data.

Notes: OLS estimation of plant and year fixed effects.

Description of plant characteristics: Multi-product and export are binary variables, Size, share of ICT to total capital, capital to employees, ratio of directly employed workers to supervisors and others are continuous variables.

Standard errors reported in the parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

In the next specification, we also include ‘Exporting Plants’ as an additional characteristic along with the previous characteristics. Sample size in this specification reduces, as data for exporting plants are available for the period from 2008–2009 to 2012–2013. Results based on this specification are reported in column (2) of Table 3. We find that exporting plants seem to have higher productivity by 2.7 per cent compared to non-exporting plants. Bernard and Jensen (2004) state that efficient plants can charge higher mark-up and overcome the costs associated with export market, thereby increasing their productivity, which leads to productivity differences across firms. The relationship between other characteristics and TFP remains qualitatively unchanged in this specification too.

Conclusion

In this article, we examine the relationship between plant-level characteristics and plant-level productivity in India using ASI data for the period from 1998–1999 to 2012–2013. In our preliminary analysis, we find that there are systematic differences between productivity levels of multi-product and single product plants as well as productivity levels of exporting and non-exporting plants. Furthermore, the number of exporting plants and multi-products plants are less than non-exporting and single product plants respectively.

From the OLS estimation with fixed effects, we observe that characteristic of plants has significant association with plant-level productivity. We find that size, in terms of employees, is positively associated with productivity. In addition, plant-level productivity is higher among plants which operate in export market or produce more than one product or invest in ICT. Moreover, it is negatively associated with capital intensity, implying that capital investment does not improve productivity. Investment in human capital, that is, employing more of skilled labour increases the productivity which is inferred by the negative relationship between the ratio of production to non-production workers and productivity.

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

Notes

References

Bernard

A. B.

Jensen

J. B.

(2004). Why some firms export. Review of Economics and Statistics , 86(2), 561–569.

Bernard

A. B.

Redding

S. J.

Schott

P. K.

(2006). Multi-product firms and product switching. American Economic Review , 100(1), 70–97.

Chatterjee

Jain

(2019). Did economic reforms affect the manufacturing productivity in India (Unpublished Manuscript). Shiv Nadar University.

Chi

(2005). Understanding the endogeneity between firm value and shareholder rights. Financial Management , 34(4), 65–76.

Chuang

Y. C.

Lin

C. M.

(1999). Foreign direct investment, R&D and spillover efficiency: Evidence from Taiwan’s manufacturing firms. The Journal of Development Studies , 35(4), 117–137.

Commander

Harrison

Menezes-Filho

(2011). ICT and productivity in developing countries: New firm-level evidence from Brazil and India. Review of Economics and Statistics , 93(2), 528–541.

Deb

A. K.

(2014). Economic reforms, capacity utilization and productivity growth in Indian manufacturing. Global Business Review , 15(4), 719–746.

De Loecker

(2011). Product differentiation, multiproduct firms, and estimating the impact of trade liberalization on productivity. Econometrica , 79(5), 1407–1451.

Dougherty

S. M.

(2009). Labour regulation and employment dynamics at the state level in India. Review of Market Integration , 1(3), 295–337.

10.

Dougherty

Herd

Chalaux

(2010). What is holding back productivity growth in India. OECD Journal: Economic Studies , 2009(1), 1–22.

11.

Goldar

Kumari

(2003). Import liberalization and productivity growth in Indian manufacturing industries in the 1990s. The Developing Economies , 41(4), 436–460.

12.

Haidar

J. I.

(2012). Trade and productivity: Self-selection or learning-by-exporting in India. Economic Modelling , 29(5), 1766–1773.

13.

Hsieh

C. T.

Klenow

P. J.

(2009). Misallocation and manufacturing TFP in China and India. The Quarterly Journal of Economics , 124(4), 1403–1448.

14.

Iyer

C. G.

(2011). Elasticity of substitution & returns to scale in Indian chemical industry. Indian Journal of Industrial Relations , 47(1), 36–42.

15.

Jorgenson

D. W.

(2001). Information technology and the US economy. American Economic Review , 91(1), 1–32.

16.

Majumdar

S. K.

(1997). The impact of size and age on firm-level performance: Some evidence from India. Review of Industrial Organization , 12(2), 231–241.

17.

Melitz

M. J.

(2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica , 71(6), 1695–1725.

18.

Oliner

S. D.

Sichel

D. E.

(2000). The resurgence of growth in the late 1990s: Is information technology the story. Journal of Economic Perspectives , 14(4), 3–22.

19.

Uchikawa

(2001). Investment boom and underutilisation of capacity in the 1990s. Economic and Political Weekly , 36(34), 3247–3254.

20.

Wagner

(2007). Exports and productivity: A survey of the evidence from firm-level data. World Economy , 30(1), 60–82.